Early prediction of university dropout: a systematic review of predictive models

Abstract

Dropout rates in higher education represent a critical challenge with academic, social, and economic implications. In Mexico, dropout rates range from 8% to 20%, affecting programs such as the Bachelor's Degree in Information Technology at the University of Guadalajara. This article presents a systematic review of predictive models applied to the early identification of student dropout, using the PRISMA approach and analyzing 15 relevant studies from databases such as Scopus, Web of Science, and ERIC. The findings highlight the use of techniques such as decision trees, neural networks, and logistic regression. However, challenges persist, such as the limited integration of psychosocial variables and the infrequent and infrequent application of predictive models throughout a student's academic career. A comprehensive methodological framework is proposed for the development of effective and ethical predictive models, with an emphasis on interdisciplinary approaches aimed at reducing dropout rates in higher education, particularly in public contexts such as that of Mexico.