Understanding the determinants of university students' academic performance has become a strategic priority for higher education institutions, especially in contexts marked by social, economic, and academic diversity. However, performance assessment remains a challenge due to the complexity of the educational process and the nonlinear nature of learning behaviors. A machine learning-based prediction model was developed using primary data from 386 university students. The performance of nine educational data mining (EDM) algorithms, including XGBoost, Random Forest, artificial neural networks (ANNs), Support Vector Machines, Decision Trees, Naive Bayes, Logistic Regression, AdaBoost, and K-Nearest Neighbors (KNN), was evaluated using demographic, socioeconomic, academic, social and family, health and wellness, infrastructure and services, and time management and extracurricular activity factors. The results reveal that machine learning is an effective tool for representing nonlinear relationships between academic performance and its determinants, allowing for accurate prediction of academic outcomes and explaining the individual contribution of each variable through sensitivity and interpretability analyses. Despite differences in their predictive accuracy, all algorithms effectively modeled educational dynamics. In particular, those models that integrate multiple student dimensions demonstrated better generalization capabilities. It is concluded that, to achieve an accurate assessment of academic performance in diverse university environments, it is essential to consider influencing factors in machine learning-based predictive models.
Yuri Reina Marín (Thu,) studied this question.