Difference between revisions of "User:Smcdougall"

Latest revision as of 20:18, 7 August 2020

About

Sarah McDougall is a senior at Bucknell University, located in Lewisburg, Pennsylvania. She is majoring in Applied Mathematical Sciences with a concentration in statistics. Upon graduation, Sarah intends to pursue a career in biomedical data science / healthcare analytics.

University Email: snm009@bucknell.edu

Weekly Work Log

Week 1

• Orientation and REU expectations
• Data Science Boot Camp (Python packages, data visualization, and classification methods)
• Initial meeting with project mentor
• Talk: Good Research Practices
• Brainstormed project goals and milestones
• Completed research on opioid abuse/dependence and opioid use disorder (OUD)

Week 2

• Completed Responsible Conduct of Research (RCR) CITI Training
• Completed Medical College of Wisconsin (MCW) Biomedical Research CITI Training
• Talk: Technical Writing
• Completed research on on statistical analyses of OUD
• Completed research on machine learning methods for predicting OUD, overdose readmissions
• Met with project group about preliminary work, next steps

Week 3

• Watched and read tutorials about scikit-learn Python package
• Completed research on machine learning methods for predicting OUD, overdose readmissions
• Met with project group to discuss datasets and exploratory data analysis
• Talk: Research Presentation by Dr. Praveen + Student Check-in
• Compiled BibTex references
• Began data preprocessing on encounter and diagnosis datasets

Week 4

• Completed data wrangling and preprocessing for hospital readmission component of project
• Began Exploratory Data Analysis (EDA) for hospital readmission component of project
• Talk: Research Presentation by Dr. Bialkowski + Student Check-in
• Planned approach for data wrangling and preprocessing for OUD component of project

Week 5

• Completed data wrangling and preprocessing for OUD component of project
• Created mini presentation on contributions made toward project so far
• Talk: Good Presentations with Dr. Brylow
• Recorded all preprocessing documentation, and narrowed down potential predictors for machine learning models
• Delivered mini presentation
• Conducted background research on opioid epidemic specifically in Wisconsin and Milwaukee County

Week 6

• Completed EDA for readmission and OUD study
• Gathered background information for final paper
• Talk: Data Ethics with Dr. Zimmer
• Conducted research on logistic regression in Python
• Conducted research on decision trees and random forest in Python
• Talk: Creating Posters with Dr. Brylow
• Tested logistic regression for both readmission and OUD study
• Completed additional background reading on imbalanced datasets, deep learning with EHR

Week 7

• Completed Random Forest, Support Vector Machine, and AdaBoost classification
• Did research on tuning hyperparameters and feature selection for classification models
• Read additional papers on deep learning with EHR (specifically RNN, LSTM, GRU)
• Completed deep learning tutorials using TensorFlow and Keras
• Added additional fields to datasets for capturing medications

Week 8

• Transformed data frames to desirable format for deep learning model
• Conducted research on RNN, LSTM, and GRU
• Retrained all models and captured prediction results in Excel sheet
• Talk: Grad School - How and Why?
• Created charts of summary statistics for final paper
• Wrote related work, data preprocessing, limitations, and future work sections of paper

Week 9

• Created data preprocessing charts, EDA tables, results table, ROC curve visualizations, confusion matrix visualizations
• Wrote background information, methods, and results sections of paper
• Talk: Industry Panel with Northwestern Mutual
• Added additional features to OUD dataset and reran all ML models
• Brainstormed information to include in poster and final presentation

Week 10

• Deliver final presentation
• Present final poster at virtual poster session
• Finish and submit final paper