TY  - JOUR
AU  - Amal El Arid
AU  - Ghalia Nassreddine
AU  - Loubna Saleh
AU  - Mohamad Al Majzoub
TI  - An Ensemble Learning Approach for Student Performance Analysis of a Higher Educational Institute using a SHAP-Based Feature Selection and Optuna Optimization
JO  - Advances in Science, Technology and Engineering Systems Journal
PY  - 2026
VL  - 11
IS  - 2
SP  - 1
EP  - 11
DO  - 10.25046/aj110201
UR  - https://www.astesj.com/v11/i02/p01/
L1  - https://www.astesj.com/?sdm_process_download=1&download_id=99076
AB  - Forecasting and assessing student performance are crucial for allowing educators to pinpoint deficiencies and promote grade improvement. A thorough comprehension of feature contributions is crucial for improving model interpretability and facilitating informed decision-making in academic institutions. Explainable artificial intelligence encompasses methodologies and strategies designed to deliver transparent and accessible rationales for the decisions rendered by artificial intelligence and machine learning algorithms. In this research paper, an interpretable gradient boosting approach for predicting student performance is introduced, including both feature selection using SHapley Additive exPlanations (SHAP)-based features and cost-sensitive decision thresholds. The proposed methodology includes hybrid resampling with SMOTE-Tomek, Optuna hyperparameter optimization with stratified cross-validation, and SHAP-guided feature selection strategy. The proposed approach is tested using a dataset of a higher educational institute in the Middle East, including student information, learning management, and a video interaction system, to make an analysis and evaluate the performance of the students. The results show both improvement of the macro F1-score and the fail-class recall by achieving an accuracy of 94%, a weighted/macro F1-score of 0.9399, and a fail-class recall of 0.9619. The suggested method facilitates trade-offs among prediction accuracy, interpretability, and fairness, bridging the divide between high-performing machine learning models and practical educational applications, hence aiding in the formulation of data-driven policies and the customization of learning experiences.
KW  - Student performance
KW  - Higher education
KW  - Ensemble learning
KW  - AI-powered tools
KW  - Explanaible AI
KW  - SHAP feature selection
KW  - Optuna optimization
ER  -