DonorsChoose is a platform where teachers request funds for their students for educational purposes. There are huge amount of proposals from teachers to be reviewed by volunteers and it's still increasing. We(with Fei Li) cleaned the data, did feature engineering(including NLP techniques to extract features from texts) and built a high-performance (AUC 0.81) predictor/classifier with Ensemble method (Gradient Boost, GRU, Random Forest, etc.). It can help to automatically choose which proposals to approve. Based on the predictor, volunteers need only to focus on the marginal cases and this greatly reduce unnecessary labor.
DonorsChoose is a platform where teachers request funds for their students for educational purposes. There are huge amount of proposals from teachers to be reviewed by volunteers and it's still increasing. We(with Fei Li) cleaned the data, did feature engineering(including NLP techniques to extract features from texts) and built a high-performance (AUC 0.81) predictor/classifier with Ensemble method (Gradient Boost, GRU, Random Forest, etc.). It can help to automatically choose which proposals to approve. Based on the predictor, volunteers need only to focus on the marginal cases and this greatly reduce unnecessary labor.