Donors Choose: Predicting Donor Excitement for Classroom Projects

Goal:

Predict whether projects listed on Donors Choose will reach full funding.

Methods:

I performed exploratory analysis and predictive modeling utilizing Python pandas and scik-itlearn on 350,000+ records of Donors Choose data, including geographical features, school type (i.e., charter, magnet, etc.), teacher program (i.e., Teach for America), project topic (i.e., books, technology, trips), requested funding amount, and funding match eligibility.

I trained multiple Naive Bayes, Logistic Regression, K-Nearest Neightbors, Bagging, and AdaBoost classifiers each tuned to a different set of hyperparameters, and evaluated them based on AUC-ROC performance to select the best model.

Files Included

• donors_choose.ipynb: My analysis, model creation, and model evaluation
• ml_explore.py: Helper functions for data exploration
• ml_pipeline_lch.py: Helper functinos for data cleaning and preprocessing
• magiclooping.py and ml_modeling.py: Helper functions for model creation and evaluation