Analysis of first-year university student dropout through machine learning models: A comparison between Universities

Abstract

The student dropout, defined as the abandonment of a high education program before obtaining thedegree without reincorporation, is a problem that affects every higher education institution. In thiswork, we used machine learning models over two Chilean universities to predict first-year engineer ing student dropout and analyze the variables that could affect the students. The results showed thatinstead of joining the datasets into a single dataset, it is better to apply a model per university. Also, the best model per dataset was gradient boosting decision trees. Finally, the interpretative models determined that a higher score in almost any test decreases the probability of dropout, being the most significant variable the mathematical test. One exception is the language test, where a higher score increases the probability of dropout.