Predictive models for the monitoring of student academic performance

Abstract

Currently, the advancement of machine learning and educational analytics technologies has enabled the development of innovative solutions in the academic field; however, many existing systems do not guarantee early risk detection or sufficient model interpretability, limiting their use in real educational settings. This study aimed to develop and implement a set of predictive models integrated into an interactive dashboard for real-time monitoring of student academic performance indicators such as grades, attendance rate, and participation score. The methodology was based on the design and comparative evaluation of regression and classification algorithms—including logistic regression, decision trees, and multilayer perceptron neural networks—applied to a dataset collected from 17 university students under controlled conditions. The optimal model achieved 95% accuracy, with a mean absolute error of less than 5% compared to grades recorded by certified educators. Finally, the study concludes that the developed system constitutes a viable, accessible, and efficient alternative for supporting pedagogical decision-making, contributing to improved preventive academic monitoring through the use of low-cost computational technologies.