Identificación y predicción de estudiantes en riesgo de deserción académica por medio de modelos basados en Machine Learning

Abstract

In the field of virtual university education in Colombia there is growing concern about the issue of student dropout, particularly in engineering schools where subjects related to natural sciences and mathematics have high academic mortality rates. This paper describes the process of identifying the most important characteristics that lead many students to abandon their studies in the Mechanics subject at the National Open and Distance University, for that there is a database between the academic periods 2018 to 2020 and On which a prediction analysis based on Machine Learning techniques was carried out, the purpose of which has been to obtain a forecast that allows identifying and anticipating possible cases of academic dropout to take the necessary measures to avoid such a situation in future cases. The pre-processing of the data and the application of the models have offered satisfactory results that allow recommendations to be made to reduce the percentage of students who drop out.