Explainable Artificial Intelligence for Student Academic Performance Prediction Using Random Forest and SHAP

Abstrak

Predicting student academic performance has become an important application of educational data mining because it enables educational institutions to identify students who require academic support at an early stage. Although machine learning algorithms have demonstrated high predictive capability, many predictive models operate as black-box systems, making it difficult for educators to understand the factors influencing prediction outcomes. This study proposes an Explainable Artificial Intelligence (XAI) framework for student academic performance prediction by integrating the Random Forest algorithm with SHapley Additive exPlanations (SHAP). The proposed methodology consists of data collection, data preprocessing, feature selection, model development, performance evaluation, and model interpretation. Random Forest was employed as the primary classification algorithm due to its robustness and high predictive performance, while SHAP was utilized to provide transparent explanations of both global and local prediction results. The experimental evaluation demonstrated that the proposed model achieved high classification performance, obtaining an accuracy of 91.67%, precision of 90.32%, recall of 93.33%, and an F1-score of 91.80%. Furthermore, SHAP analysis identified Previous GPA, Final Examination Score, Attendance, Assignment Score, and Study Hours as the most influential factors affecting student academic performance. The integration of Random Forest and SHAP not only improves prediction reliability but also enhances model transparency by explaining the contribution of each feature to prediction outcomes. Consequently, the proposed framework supports evidence-based academic decision-making, facilitates early identification of at-risk students, and provides educators with interpretable insights for designing effective academic intervention strategies. These findings demonstrate that Explainable Artificial Intelligence can significantly improve the practical applicability and trustworthiness of machine learning models in educational environments.