ARCHIVES

Early identification of students at risk of poor academic performance is essential for improving learning outcomes and reducing dropout rates in higher education institutions. The increasing adoption of digital learning platforms has resulted in the generation of large-scale educational data, enabling the application of machine learning techniques for predictive analytics. Although existing studies report high prediction accuracy, most machine learning–based models operate as black boxes, limiting their interpretability and practical adoption by educators. To address this limitation, this paper proposes an explainable machine learning framework for early prediction of student academic performance. The proposed approach employs a Random Forest classifier for accurate risk prediction and integrates SHapley Additive exPlanations (SHAP) to provide both global and local interpretability of model decisions. Experiments conducted on the UCI Student Performance dataset demonstrate that the proposed framework achieves an accuracy of 91.14%, an F1-score of 86.79%, and an ROC–AUC of 0.9655, outperforming baseline classifiers. Furthermore, SHAP-based explanations identify key academic and behavioral factors influencing student performance and provide individualized explanations for at-risk students. The results confirm that the proposed framework offers both high predictive performance and transparency, making it suitable for real-world educational decision-support systems