The goal of this project is to model undergraduate student performance as measured by grades in STEM courses. I approached this as both a classification and regression problem.

The project primarily relies on Stata for data cleaning and the Python libraries scikit-learn for the learning algorithms and imbalanced-learn to address class imbalance.

Inputs: Raw administrative data from several DB's. Outputs: Predicted STEM course grades based on available data.

This work is ongoing, and this repo is not updated live.

The ./data-cleaning/ folder contains all Stata .do files used to clean the training data. The entire procedure can be completed by running the ./data-cleaning/0.0_data_cleaning_control.do file.

The ./src/ folder contains the Python code used to train and test various learning algorithms for our problem. This procedure can be completed by running the ./src/master.py file.

conda create --prefix=$HOME/myenvs/nsfstem python=3.10
source activate $HOME/myenvs/nsfstem
pip install -r PATHTOPROJECTFOLDER/requirements.txt
pip install --upgrade gensim

I attended the Association for Education Finance and Policy 2024 conference where I presented progress in this work. We received great feedback and will be including additional analyses and modelling strategies.