EduBoost: A High-Accuracy XGBoost Framework for Student Academic Performance Prediction

Resumen

The study presents EduBoost, a high-accuracy machine learning framework based on the XGBoost algorithm for predicting student academic performance. Conventional assessment approaches are often retrospective and fail to identify academically at-risk students at an early stage, thereby limiting timely and targeted educational interventions. EduBoost addresses the limitation by formulating the prediction task as an ordinal multi-class classification problem, in which final academic grades are discretized into low, medium, and high performance categories. Extensive experiments are conducted on a real-world student performance dataset comprising demographic, socio-economic, academic, and behavioral attributes. The proposed framework achieves an overall classification accuracy of 95%, consistently outperforming baseline models such as logistic regression, decision trees, support vector machines, and random forests. Moreover, feature importance analysis provides interpretable insights into the key factors influencing academic outcomes, enhancing transparency and decision support. The experimental results demonstrate that EduBoost is a robust, accurate, and interpretable solution that can effectively support early academic risk identification, personalized learning strategies, and data-driven decision-making in educational institutions.