Machine Learning algorithms to predict desertion in the faculty of Engineering Sciences at the Corporación Universitaria Antonio José de Sucre.

Abstract

Desertion can be understood as the withdrawal and subsequent abandonment of school activity by students, due to different family, economic, social and other factors. All educational levels are affected by this scourge, being the university one of the most affected, especially during this time of pandemic caused by the SARS-CoV 19 virus. Due to the complexity of this problem, and the great impact it has generated at the educational level, many universities have proposed different intervention strategies to reduce dropout rates. The difficulty is that many of these strategies lack effectiveness, since they do not take into account the different causes, which are different for each case. On the other hand, it is important to have accurate and reliable information to determine the population in order to identify possible cases of dropout, and to take preventive actions to reduce the student dropout rate. In this sense, the purpose of this study is to analyze different predictive algorithms based on Machine Learning that enable early detection of possible cases of desertion in the Faculty of Engineering at the Corporación Universitaria Antonio José De Sucre. Among such algorithms we can mention decision trees, logistic regression and support vector machines. These will be trained with historical data and then tested to determine their performance, to finally choose the most appropriate within the context of the current problem.