https://reu.cs.mu.edu/api.php?action=feedcontributions&user=Jaired&feedformat=atomREU@MU - User contributions [en]2024-03-29T15:51:02ZUser contributionsMediaWiki 1.23.13https://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-08-02T14:41:45Z<p>Jaired: /* Week 10 (30 Jul - 3 Aug) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
* Ran co-researcher through code since I'm forsaking him soon<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
* Worked on paper<br />
* Went through code with co-researcher<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
* Final project meeting<br />
* Worked on paper more<br />
* Ran a final experiment<br />
<br />
== Thursday ==<br />
* Worked on paper<br />
* Wrote readmes<br />
* Created a directory with my work for GasDay<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-31T15:46:42Z<p>Jaired: /* Thursday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
* Ran co-researcher through code since I'm forsaking him soon<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
* Worked on paper<br />
* Went through code with co-researcher<br />
* Wrote some readmes<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
* Final project meeting<br />
* Worked on paper more<br />
* Wrote some more readmes<br />
<br />
== Thursday ==<br />
* REU presentations<br />
* Worked on paper<br />
* Wrote readmes<br />
* Created a directory with my work for GasDay<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-31T15:46:07Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
* Ran co-researcher through code since I'm forsaking him soon<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
* Worked on paper<br />
* Went through code with co-researcher<br />
* Wrote some readmes<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
* Final project meeting<br />
* Worked on paper more<br />
* Wrote some more readmes<br />
<br />
== Thursday ==<br />
* REU presentations<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-31T15:45:44Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
* Ran co-researcher through code since I'm forsaking him soon<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
* Worked on paper<br />
* Went through code with co-researcher<br />
* Wrote some readmes<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
<br />
== Thursday ==<br />
* REU presentations<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-24T11:50:33Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
* Ran co-researcher through code since I'm forsaking him soon<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
<br />
== Thursday ==<br />
* REU presentations<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-24T11:49:59Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
<br />
== Thursday ==<br />
* REU presentations<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-23T18:51:32Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Paper work<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
<br />
== Thursday ==<br />
* REU presentations<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-23T18:35:30Z<p>Jaired: /* Monday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
<br />
== Thursday ==<br />
* REU presentations<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-23T15:00:54Z<p>Jaired: /* Monday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum at AMU<br />
* Talked to random people about my poster<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
<br />
== Thursday ==<br />
* REU presentations<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-23T14:59:11Z<p>Jaired: /* Week 10 (30 Jul - 3 Aug) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
<br />
== Monday ==<br />
* Poster forum<br />
* Worked on paper<br />
<br />
== Tuesday ==<br />
* REU presentation preparation<br />
<br />
== Wednesday ==<br />
* REU presentations and lunch<br />
<br />
== Thursday ==<br />
* REU presentations<br />
<br />
== Friday ==<br />
* Take me home<br />
* country roads<br />
* to the place<br />
* I belong</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-23T14:57:55Z<p>Jaired: /* Week 9 (23 Jul - 27 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
<br />
== Monday ==<br />
* Operator error on time horizon stop, 3 instead of 30<br />
* Ran script on Dr. Corliss's machine<br />
* Converted paper to Elsevier format<br />
* Worked on paper<br />
* Project meeting<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Tuesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Wednesday ==<br />
* Paper work<br />
* Worked on GasDay presentation for seminar<br />
<br />
== Thursday ==<br />
* Finishing touches on GasDay presentation<br />
* Presentation at GasDay<br />
* GasDay seminar<br />
* Weekly REU lunch meeting<br />
* Project meeting<br />
* Worked on paper<br />
<br />
== Friday ==<br />
* Paper work<br />
<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-20T14:25:00Z<p>Jaired: /* Friday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Created sensitivity figures for multiple time horizons<br />
* Left a script running for the weekend to produce figures<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-18T13:47:06Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created MATLAB objects to store models and model parameters<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Work on paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-18T11:53:45Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Tested out adding a new feature to see if prediction is a little more accurate<br />
* Created an alarm forecasting script<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Work on paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-18T11:52:58Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Work on paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-17T17:03:38Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Created a forecasting script independent of training<br />
* Worked on alarm prediction script<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
* Meeting with Dr. Povinelli<br />
<br />
== Wednesday ==<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Work on paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-17T14:22:57Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* More scrutinizing of code, determined inaccuracies are caused by a bias in the algorithm<br />
* "Cleaned" some anomalous spikes in data<br />
* Created a training function<br />
* Created a forecasting script independent of training<br />
* Worked on alarm prediction script<br />
* Tried to make the poster a little prettier<br />
* Generated some figures for poster<br />
<br />
== Wednesday ==<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Work on paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-17T14:20:38Z<p>Jaired: /* Monday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Scrutinized ensemble of direct forecasts script for bugs<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* Alarm forecasting<br />
<br />
== Wednesday ==<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Work on paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-16T17:09:04Z<p>Jaired: /* Monday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Finding lag bug in ensemble of direct forecast script<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
== Tuesday ==<br />
* Alarm forecasting<br />
<br />
== Wednesday ==<br />
* Poster work<br />
<br />
== Thursday ==<br />
* Poster work<br />
* GasDay Seminar<br />
* REU Working Lunch<br />
* Project meeting<br />
<br />
== Friday ==<br />
* Finishing touches on poster<br />
* Work on paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-16T13:13:43Z<p>Jaired: /* Week 9 (23 Jul - 27 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Finding lag bug in ensemble of direct forecast script<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-16T13:13:29Z<p>Jaired: /* Week 8 (16 Jul - 20 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
== Monday ==<br />
* Finding lag bug in ensemble of direct forecast script<br />
* Looked for a target publication venue<br />
* Worked on method section in research paper<br />
<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-13T19:21:58Z<p>Jaired: /* Friday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast models<br />
* Direct forecast testing and tuning<br />
* Reorganized git repository<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-13T16:42:04Z<p>Jaired: /* Friday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast methods<br />
* Direct forecast testing and tuning<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-13T13:46:54Z<p>Jaired: /* Friday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Worked on research paper<br />
* Went over very rough draft of paper with Dr. Corliss<br />
* Ensembled new direct forecast methods<br />
* Predicting actual alarms<br />
* Direct forecast test<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-12T18:29:52Z<p>Jaired: /* Thursday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with several different time horizons<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Work on research paper<br />
* Additional tests with LS-SVM on Dr. Corliss's machine<br />
* Predicting actual alarms<br />
* Direct forecast test<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-12T13:25:31Z<p>Jaired: /* Thursday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Creation of naive model and its RMSE and MAE<br />
* Direct forecast with 30 minute time horizon<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Work on research paper<br />
* Additional tests with LS-SVM on Dr. Corliss's machine<br />
* Predicting actual alarms<br />
* Direct forecast test<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-11T17:34:38Z<p>Jaired: /* Week 7 (9 Jul - 13 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Applying LS-SVM model to forecast alarm data<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
== Friday ==<br />
* Work on research paper<br />
* Additional tests with LS-SVM on Dr. Corliss's machine<br />
* Predicting actual alarms<br />
* Direct forecast test<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-11T17:34:04Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Tweaked SVM and SVR figures<br />
* Worked on Methods section for research paper<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Applying LS-SVM model to forecast alarm data<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-11T14:44:53Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Studied SVMs and SVRs heavily to be able to explain them in the research paper<br />
* Worked on the background for research paper<br />
* Additional tests with LS-SVM on Dr. Corliss's machine<br />
* Direct forecast test<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Applying LS-SVM model to forecast alarm data<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-11T12:06:58Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
* Created figures to explain SVM and SVR<br />
<br />
== Wednesday ==<br />
* Work on research paper<br />
* Additional tests with LS-SVM on Dr. Corliss's machine<br />
* Direct forecast test<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Applying LS-SVM model to forecast alarm data<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-10T16:42:37Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Worked on background of research paper<br />
<br />
== Wednesday ==<br />
* Work on research paper<br />
* Additional tests with LS-SVM on Dr. Corliss's machine<br />
* Direct forecast test<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Applying LS-SVM model to forecast alarm data<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/Summer_2018_ProjectsSummer 2018 Projects2018-07-10T14:18:01Z<p>Jaired: </p>
<hr />
<div>'''[[Utilizing Graph-Cuts in Image Segmentation to Isolate the Heart within CT Scans]]''':<br />
<br />Student: [[User:saweiner|Sam Weiner]].<br />
<br />Mentor: Dr. Naveen Bansal<br />
<br />
<br />
'''[[Intrusion Detection in Swarm Robotics]]''':<br />
<br />Student Researcher: [[Lindsey Coffee-Johnson]]<br />
<br />Mentor: Dr. Debbie Perouli<br />
<br />
<br />
'''[[SUPREME: A Cancer Subtype Prediction Methodology by Integrating High-Dimensional Biological Datasets]]''':<br />
<br />Student Researcher: [[User:Jsu|Jeanne Su]].<br />
<br />Mentor: [http://www.marquette.edu/mscs/facstaff-bozdag.shtml Dr. Serdar Bozdag].<br />
<br />
<br />
'''[[A Qualitative Study of Wisconsin Computer Science in K-12]]''':<br />
<br />Student Researchers: [[User:Djeffers|Darren Jefferson]], [[User:Pmoras|Peter Moras]]<br />
<br />Mentor: [[User:Brylow|Dr. Dennis Brylow]]<br />
<br />
<br />
'''[[Expanding Curricula for Parallel Computing Fundamentals in Computer Architecture and Hardware Courses]]''':<br />
<br />Student Researcher: [[User:Blevando|Benjamin Levandowski]].<br />
<br />Mentor: [[User:Brylow|Dr. Dennis Brylow]].<br />
<br />
<br />
'''[[Multicore Embedded Operating System Components for Education]]''':<br />
<br />Student Researcher: [[User:Bweither|Brian Weithers]].<br />
<br />Mentor: [[User:Brylow|Dr. Dennis Brylow]].<br />
<br />
<br />
'''[[Deconstructing Spatial Clustering Algorithms To Explore Biases in Crime Analysis]]''':<br />
<br />Student Researchers: [[User:KWeathington|Katy Weathington]], [[User:Lauraschultz|Laura Schultz]], [[User:Quinci Henry|Quinci Henry]].<br />
<br />Mentor: Dr. Shion Guha<br />
<br />
<br />
'''[[GasDay Alarm Prediction]]''':<br />
<br>Student Researcher: [[User:jaired|Jaired Collins]]<br />
<br>Mentors: [http://www.marquette.edu/electrical-computer-engineering/povinelli-richard.php Dr. Richard Povinelli], [http://www.marquette.edu/mscs/facstaff-corliss.shtml Dr. George Corliss]<br />
<br />
<br />
'''[[Gaussian Stochastic Processes for Coupling Landslide Hazards]]''':<br />
<br /> Student Researcher: [[User:lwebster|Lindsay Webster]]<br />
<br /> Mentor: [http://www.marquette.edu/mscs/facstaff-spiller.shtml Dr. Elaine Spiller]<br />
<br />
<br />
'''[[Forecasting Stock Prices using Social Media Analysis]]''':<br />
<br />Student Researcher: [[User:Dalmeidad|Dawson d'Almeida]]<br />
<br />Mentor: Dr. Praveen Madiraju</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-10T14:13:52Z<p>Jaired: /* Week 7 (9 Jul - 13 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Povinelli to go over revised material for his trip<br />
* Meeting with Dr. Corliss to get his opinion on prepared items<br />
* Work on research paper<br />
<br />
== Wednesday ==<br />
* Work on research paper<br />
* Additional tests with LS-SVM on Dr. Corliss's machine<br />
* Direct forecast test<br />
<br />
== Thursday ==<br />
* Work on research paper<br />
* Applying LS-SVM model to forecast alarm data<br />
* GasDay seminar<br />
* REU Working Lunch<br />
<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-10T14:10:25Z<p>Jaired: /* Week 7 (9 Jul - 13 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Another meeting with Dr. Povinelli to go over his and the graduate student's trip to the customer company<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-09T17:19:23Z<p>Jaired: /* Week 7 (9 Jul - 13 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
<br />
== Monday ==<br />
* Meeting with Dr. Povinelli to discuss project and his and graduate student's visit to the customer company<br />
* Created different figure of LS-SVM to only include the 30 minute forecast value every minute<br />
* Analyzed all sensor information from all the sites to find anomalies<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-05T15:50:27Z<p>Jaired: /* Thursday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* Started working on a programmatic rule-based ensemble of differently trained models<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-05T15:49:51Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Manual ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-05T14:45:26Z<p>Jaired: /* Thursday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Rule-based ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to use more training data<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-05T14:40:53Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Rule-based ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to using more training data<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-05T14:40:35Z<p>Jaired: /* Thursday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Remote desktop into a better computer to using more training data<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-04T18:32:19Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Created several figures to show LS-SVM for regression<br />
<br />
== Thursday ==<br />
* Direct approach<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-04T12:51:21Z<p>Jaired: /* Week 6 (2 Jul - 6 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of troughed data with LS-SVM<br />
* Literature search on time series/forecast*/machine learning<br />
<br />
== Wednesday ==<br />
* Created "artificial" data with extra peaks in it to improve accuracy of peaked data with LS-SVM<br />
* Ensemble of peaked, trough, and normal LS-SVM models for forecasting<br />
* Worked on research paper<br />
* Literature search<br />
<br />
== Thursday ==<br />
* Direct approach<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-04T11:46:16Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of peaked data with LS-SVM<br />
<br />
== Wednesday ==<br />
* Ensemble of peak trained and normal trained iterative approach to LS-SVM forecasting<br />
* Worked on research paper<br />
* Literature search<br />
<br />
== Thursday ==<br />
* Created data with more peaks to train LS-SVM to predict them better<br />
* Ensemble of peaked LS-SVM and more normal LS-SVM<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-03T18:11:03Z<p>Jaired: /* Thursday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of peaked data with LS-SVM<br />
<br />
== Wednesday ==<br />
* Started direct approach instead of iterative forecasting<br />
* Worked on research paper<br />
* Literature search<br />
* Ensemble of direct vs. iterative forecast<br />
<br />
== Thursday ==<br />
* Created data with more peaks to train LS-SVM to predict them better<br />
* Ensemble of peaked LS-SVM and more normal LS-SVM<br />
* REU working lunch with Dr. Corliss<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-03T18:10:10Z<p>Jaired: /* Wednesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of peaked data with LS-SVM<br />
<br />
== Wednesday ==<br />
* Started direct approach instead of iterative forecasting<br />
* Worked on research paper<br />
* Literature search<br />
* Ensemble of direct vs. iterative forecast<br />
<br />
== Thursday ==<br />
* Created data with more peaks to train LS-SVM to predict them better<br />
* Ensemble of peaked LS-SVM and more normal LS-SVM<br />
* REU working lunch<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-03T18:09:38Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Created "artificial" data with extra troughs in it to improve accuracy of peaked data with LS-SVM<br />
<br />
== Wednesday ==<br />
* Worked on research paper<br />
* Literature search<br />
* Ensemble of direct vs. iterative forecast<br />
<br />
== Thursday ==<br />
* Created data with more peaks to train LS-SVM to predict them better<br />
* Ensemble of peaked LS-SVM and more normal LS-SVM<br />
* REU working lunch<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-03T15:42:40Z<p>Jaired: /* Week 6 (2 Jul - 6 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Coded new direct forecast instead of iterative<br />
<br />
== Wednesday ==<br />
* Worked on research paper<br />
* Literature search<br />
* Ensemble of direct vs. iterative forecast<br />
<br />
== Thursday ==<br />
* Created data with more peaks to train LS-SVM to predict them better<br />
* Ensemble of peaked LS-SVM and more normal LS-SVM<br />
* REU working lunch<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-03T15:40:35Z<p>Jaired: /* Tuesday */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Implemented long short-term memory recurrent neural network, peaks too often<br />
* Tried several configurations of LS-SVM NAR<br />
* Coded new direct forecast instead of iterative<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-03T12:18:34Z<p>Jaired: /* Week 6 (2 Jul - 6 Jul) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
<br />
== Monday ==<br />
* Confirmed some math with Dr. Corliss<br />
* Messing around with LS-SVM to get it to work<br />
* Created normalized time series<br />
* Extracted correlated information from multiple sensors<br />
<br />
== Tuesday ==<br />
* Work on long short-term memory recurrent neural network<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jairedhttps://reu.cs.mu.edu/index.php/User:JairedUser:Jaired2018-07-03T12:17:01Z<p>Jaired: /* Week 5 (25 Jun - 29 Jun) */</p>
<hr />
<div>I am Jaired Collins, a student at Missouri Southern State University in Joplin, MO, currently majoring in computational mathematics. Outside of academia, I enjoy cycling, weights, table tennis, tennis, and cooking.<br />
<br />
= '''Week 1 (29 May - 1 Jun)''' =<br />
== Tuesday ==<br />
* Initial meeting to get introduced into the program.<br />
* Lunch meeting with mentors.<br />
* Obtained MU and MSCS account logins and ID card<br />
* Paperwork<br />
<br />
== Wednesday ==<br />
* Met with Dr. Corliss to get acquainted with the GasDay lab and receive a key.<br />
* Met with Dr. Povinelli and another student to set up initial parameters of the project.<br />
* Literature review.<br />
<br />
== Thursday ==<br />
* More literature review<br />
* Seminar with GasDay<br />
* Meeting with Dr. Povinelli and student to establish a weekly plan and talk about the project.<br />
<br />
== Friday ==<br />
* Literature review<br />
* Planned for a presentation at a GasDay Seminar on the 21st of June<br />
* Data cleaning<br />
<br><br />
<br />
= '''Week 2 (4 Jun - 8 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli and Colin<br />
* Literature search<br />
* Cleaned up more data<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Data cleaning with bash and Matlab<br />
* Played around with data and SVMs in MatLab<br />
* Meeting with research consultant<br />
* Literature review<br />
<br />
== Wednesday ==<br />
* Cross validation with k-Fold<br />
* GasDay Camp<br />
* Literature search<br />
<br />
== Thursday ==<br />
* More built-in SVM work in MatLab<br />
* GasDay camp<br />
* REU working lunch meeting, bad presentation<br />
<br />
== Friday ==<br />
* Meeting with Dr. Corliss<br />
* Creation of ShareLaTeX project<br />
* Converted temporal pressure data into nonuniform time series data<br />
* Tried resampling time series data into uniform data<br />
<br><br />
<br />
= '''Week 3 (11 Jun - 15 Jun)''' =<br />
== Monday ==<br />
* Meeting with Dr. Corliss<br />
* Meeting with Dr. Povinelli<br />
* Resampled nonuniform data<br />
* Forecasting with ARIMA<br />
* Left computer running to finish ARIMA models<br />
<br />
== Tuesday ==<br />
* Examined and saved ARIMA models<br />
* Dissimilarity-based approach to prediction, where correlation coefficients between sensors are used as features<br />
* SVM and k-fold cross validation at k = 5, 10, and 20<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss to elucidate LS-SVM for regression<br />
* Coding LS-SVM<br />
* RCR lunch meeting<br />
<br />
== Thursday ==<br />
* Meeting with Dr. Corliss, further explained LS-SVM<br />
* GasDay seminar<br />
* Coding of LS-SVM for regression<br />
* Meeting with Dr. Povinelli<br />
* Called contact to clarify threshold values<br />
<br />
== Friday ==<br />
* Out for wedding<br />
<br />
<br><br />
<br />
= '''Week 4 (18 Jun - 22 Jun)''' =<br />
== Monday ==<br />
* Out for wedding<br />
<br />
== Tuesday ==<br />
* Wrote working abstract<br />
* Wrote summaries for related works section<br />
* Read Anomaly Detection in Streaming Non-stationary Temporal Data<br />
* Read Chapter 5 of Operations & Maintenance Best Practices Release 3.0<br />
* Tested different features in LS-SVM; residuals are very low<br />
<br />
== Wednesday ==<br />
* Meeting with Dr. Corliss, showed LS-SVM results<br />
* Work on conference paper with graduate student<br />
* LS-SVM Tuning<br />
* Work on PPT for presentation<br />
<br />
== Thursday ==<br />
* Practice presentation at GasDay Seminar<br />
* Working REU lunch<br />
* Knowledge and Information Discovery with Dr. Povinelli<br />
* Project meeting with Dr. Povinelli and graduate student<br />
<br />
== Friday ==<br />
* Recursive nonlinear autoregressive model with LS-SVM<br />
* Clarification on building different models with different time horizons instead of recursive approach<br />
* Coding nonlinear autoregressive model with LS-SVM with multiple time horizons<br />
* Work on abstract<br />
* Work on background<br />
<br><br />
<br />
= '''Week 5 (25 Jun - 29 Jun)''' =<br />
== Monday ==<br />
* Worked on slides for presentation on Thursday<br />
* Tried to debug direct multiple time horizon least squares support vector machine nonlinear autoregressive model (LS-SVM NAR)<br />
* Showed abstract to Dr. Povinelli, got feedback<br />
* Resampled data to 1 minute and wrote several MATLAB scripts to automate the LS-SVM NAR process<br />
* Came to the conclusion that LS-SVM NAR is not predicting well at all. Plotted values were erroneous and coincidentally showed promising results<br />
<br />
== Tuesday ==<br />
* Meeting with Dr. Corliss<br />
* Read more into SVMs and SVRs to help the LS-SVM NAR algorithm<br />
* Implementation of radial basis function kernel for LS-SVM NAR<br />
<br />
== Wednesday ==<br />
* Debugging LS-LSVM NAR with RBF kernel with test cases<br />
* Tuning LS-LSVM NAR hyperparameters<br />
* Presentation preparation<br />
<br />
== Thursday ==<br />
* More preparation for REU Midterm<br />
* GasDay Seminar<br />
* Midterm presentations<br />
* KID Seminar<br />
* Meeting with Dr. Povinelli to discuss project<br />
<br />
== Friday ==<br />
* Put a lot of effort into making LS-SVM work<br />
* Meeting with Dr. Povinelli to help with LS-SVM and things to do while he is on vacation<br />
<br><br />
<br />
= '''Week 6 (2 Jul - 6 Jul)''' =<br />
Generate new ideas for algorithm<br />
<br><br />
<br />
= '''Week 7 (9 Jul - 13 Jul)''' =<br />
Code and test algorithm<br />
<br><br />
<br />
= '''Week 8 (16 Jul - 20 Jul)''' =<br />
More experiments with algorithm<br />
<br><br />
<br />
= '''Week 9 (23 Jul - 27 Jul)''' =<br />
Work on final draft of research paper<br />
<br>Work on poster<br />
<br><br />
<br />
= '''Week 10 (30 Jul - 3 Aug)''' =<br />
Prepare final presentation</div>Jaired