Construcción de modelos de machine learning con aprendizaje supervisado para determinar la deserción académica en estudiantes universitarios

Abstract

In this investigation, the objective has been set to make use of machine learning to generate predictions about the status of the student dropout, specifically for the Electrical Engineering and the Cadastral Engineering programs of the District University of Francisco José de Caldas. For the case of Electrical Engineering, data of 1834 students were used from semester 2009-1 to the semester 2018-3, while Cadastral Engineering, data of 2335 students were used from semester 2009-3 to the semester 2018-3. Features related to pre-university, socioeconomic, demographic, academic, and institutional factors are found in the databases. To make the predictions in the 10 semesters of duration of both programs, the following algorithms were used: decisión tree, logistic regression, KNN (K-nearest neighbor), SVM (Support Vector Machine) classifier and Naive Bayes. With the results obtained, it has been concluded that machine learning it’s a good option to predict the student dropout and that the information obtained by the predictions could be useful to help in the search of strategies that allow to reduce the universitary dropout.