DonorsChoose.org is a crowdfunding website that connect public school teachers to donors on classroom projects. I would like to build a tool for new teachers to predict whether their first classroom project will get funded on the website.
I implemented logistic regression, K-nearest Neighbors, Random Forest and Gradient Boosting classifiers to predict the crowdfunding success probability of new classroom projects on DonorsChoose.org.
My final model is a gradient boosting classifier model with Area Under the ROC Curve (AUC) score of 0.718.
| Model | AUC |
|---|---|
| Gradient Boosting + KNN | 0.718 |
| Random Forest + KNN | 0.711 |
| Logistic Regression + Select K Best + KNN | 0.664 |
More details are coming soon!
See the LICENSE file for license rights and limitations (MIT).