Academic Performance Prediction For High-Education Students Using Machine Learning

This study evaluates machine learning algorithms for predicting academic performance in higher education, utilizing a dataset of 1,259 Computer Engineering student records from the University of Tras-os-Montes and Alto Douro (UTAD) spanning from 2011 to 2023. Incorporating demographic factors, academic trajectory, and performance metrics, the study prioritizes first-year data to assess the feasibility of early detection. We compare Artificial Neural Networks (ANN), Decision Trees (DT), and Random Forest (RF) for multi-class performance classification and binary degree completion prediction. Results show tree-based models achieve superior performance, with Decision Trees reaching 0.81 accuracy in multi-class classification and 0.90 accuracy in predicting degree completion. These findings show that machine learning-based prediction systems can effectively support early identification of at-risk students and enable timely pedagogical interventions.