The integration of explainable AI in Educational Data Mining for student academic performance prediction and support system

The academic performance of students is important because it is a crucial indicator of their educational life. The absence of prior experience, guidelines, and social and personal factors can contribute to unforeseen issues that adversely impact a student’s academic performance. An Artificial Intelligence (AI)-driven Educational Data Mining (EDM) system can serve as a solution in this regard. In this study, we proposed an EDM system that integrates machine learning to classify student academic performance and explainable AI techniques for explainability. We utilize standard ML models such as Decision Tree and Random Forest, and Boosting models such as Gradient Boosting and Extreme Gradient Boosting (XGB) for multiclass classification, and SHAP, Shapash, Eli5, and LIME as XAI techniques for both global and local explainability. For the student’s academic performance classification task, the XGB model outperforms other models and achieved 83% accuracy. The analysis indicates that the evaluations of the curricula, courses, and economic variables of different semesters greatly impact the student’s academic performances. To provide proper direction to students, we also developed a support system that uses the trained model from our analysis and provides recommendations and suggestions based on the input data. The further study integrates real-time data and mobile applications for the support system. • An ML-enabled EDM approach is proposed to predict student academic performance. • XAI methods are used to find the key factors affecting student academic performance. • Decision support system offers personalized support based on real-time data.