Predicting Student Academic Success through Explainable Machine Learning Models: A Comparative Study of BRF, XGBoost, and CatBoost

Abstract

This study presents an explainable machine learning framework for predicting academic success among university students using Balanced Random Forest (BRF), XGBoost, and CatBoost. The dataset, collected from undergraduate computer science students in Bangladesh, encompasses academic, behavioral, and socio-demographic attributes. To handle class imbalance, SMOTE-based oversampling was applied, and model thresholds were optimized using the F1-score. Model interpretability was achieved using SHAP (SHapley Additive exPlanations), enabling transparent identification of influential features such as attendance, study habits, income, and academic progression. Among all models, CatBoost achieved the highest macro F1-score and ROC-AUC, demonstrating its robustness in handling heterogeneous educational data. The results highlight how interpretable ensemble learning can support early risk detection and data-driven academic interventions, thereby bridging the gap between predictive performance and educational transparency.