Enhancing Student Performance Prediction Through Ensembles of Machine Learning Models and Explainable Artificial Intelligence

Predicting student performance accurately is essential for personalizing education, allocating resources, and helping at-risk students. This study explores application of Machine Learning (ML) models and Explainable Artificial Intelligence (XAI) methods techniques such as Local Interpretable Modelagnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP), to enhance the interpretability of predictions in educational settings. By utilizing student demographic and behavior data, we built ML models, specifically BaggingClassifier ensemble models by achieving an Accuracy of 0.95, ROC AUC of 0.97, and F1 Score of 0.95, to predict student performance and also compare their performance with other single models. Additionally, LIME and SHAP were applied on the best performing model to determine the most influencer features on the prediction. Specifically, our analysis revealed that StudentAbsenceDays and VisITedResources are most influencial features. The applied bagging ensemble method not only improved the accuracy of the predictions but also highlight the features contributed to the prediction, which can ultimately aid academicians to make better decisions for the students.