A machine learning approach to enrollment prediction in Chicago Public School

Abstract

Chicago Public School (CPS) allocates billions of dollars to hundreds of public schools including new schools in its system based on prediction of enrollment number for the next year, of which ninth grade is most difficult to be projected. In this project, we propose a method called conditional logistic regression to help them improve the enrollment projection on new high schools. We also design an ensemble model to predict ninth grade. Based on our experiments on two years, our method is better than the current method of CPS.