Revisión Sistemática: Machine Learning en la Predicción de Deserción Académica

The goal of this work was to elaborate a systematic review on the use of Machine Learning in the prediction of academic dropout, for which 4 research questions were raised and using the methodology known as PRISMA, 30 articles were found that passed the proposed exclusion and inclusion criteria, and the Zotero software (Bibliographic Reference Manager) was used to facilitate the management of the articles found. The results expressed that Machine Learning can be applied in distinct sectors such as financial and educational, also that the most used variables to predict academic dropout were the student’s sex (60%), grade point average (53.33%), age (43.33%), place of origin (33.33%), type of school of origin (26.67%), employment situation (23.33%) and marital status (23.33%), finally, it was evidenced that the most common techniques in the prediction of academic dropout were those of Decision Tree (66.67%) and Artificial Neural Network (50%), and finally, it is evident that the most common metrics in the evaluation of predictive models of academic dropout are precision (63.33%), accuracy (60%), and sensitivity (50%).