Dropout prediction and prevention

Alvaro Quezada Hofflinger and I have built models to predict the risk of dropout at a Chilean university. Our predictions become much more accurate after students’ first semester in college. Some of the most predictive variables–such as whether a student received a fellowship or was admitted to their first choice major–are under the university’s control.

lift curves

I have also blogged about the broad potential of predictive analytics in higher education: