An Explainable Ensemble Learning Approach for Student Performance Prediction
TL;DR
An explainable ensemble learning approach for student performance prediction is proposed by integrating Random Forest, XGBoost, and LightGBM classifiers with SHapley Additive exPlanations (SHAP) to improve both predictive performance and model transparency.
Abstract
Student performance prediction has become an important research area in educational data mining because it enables educational institutions to identify academically at-risk students and implement timely intervention strategies. Although machine learning techniques have significantly improved prediction accuracy, many existing models function as black-box systems that provide limited explanation of the factors influencing academic performance. The absence of interpretability restricts educators from understanding the reasoning behind prediction outcomes and reduces confidence in adopting artificial intelligence-based decision support systems. This study proposes an explainable ensemble learning approach for student performance prediction by integrating Random Forest, XGBoost, and LightGBM classifiers with SHapley Additive exPlanations (SHAP). The proposed framework includes data preprocessing, feature engineering, ensemble learning, and feature interpretation to improve both predictive performance and model transparency. A publicly available student performance dataset containing 14,003 student records with 16 academic, behavioural, and demographic attributes is used for experimental evaluation. The dataset is divided into training and testing subsets using an 80:20 ratio. The performance of the proposed approach is evaluated using Accuracy, Precision, Recall, F1-Score, and ROC-AUC and compared with conventional machine learning models. SHAP analysis is employed to identify the contribution of individual features influencing student performance, enabling transparent and interpretable predictions. The proposed approach assists educators in identifying the key factors affecting academic achievement and supports timely intervention for improving student success. The results demonstrate that combining ensemble learning with explainable artificial intelligence provides an effective and reliable framework for educational decision-making.