Predicting K-12 Dropout

Abstract

Dropout remains a persistent challenge within high school education. In this paper, we present a case study on automatically detecting whether a student is at-risk of dropout within a diverse school district in Texas. We predict whether a student will drop out in a future school year from data on students’ discipline, attendance, course-taking, and grades, using a logistic regression framework. We discuss the predictive properties of the model, and the features that are predictive of dropout in this context.