Explainable Artificial Intelligence-Based Early Warning System for Student Dropout Prediction

Abstract

Student dropout remains one of the major challenges faced by higher education institutions because it negatively affects academic performance, institutional reputation, and educational sustainability. Early identification of students at risk of dropping out enables universities to implement timely interventions that improve student retention. Although machine learning models have demonstrated promising predictive performance, many of these approaches operate as black-box systems, limiting their transparency and practical adoption in educational decision-making. Therefore, this study proposes an Explainable Artificial Intelligence (XAI)-Based Early Warning System for student dropout prediction by integrating the XGBoost classification algorithm with SHapley Additive exPlanations (SHAP). The proposed framework consists of six stages: data collection, data preprocessing, feature engineering, machine learning model development, explainability analysis, and performance evaluation. To demonstrate the proposed methodology, a synthetic dataset representing higher education student records was utilized. The simulated experimental results showed that the proposed model achieved an Accuracy of 89.5%, Precision of 84.0%, Recall of 82.0%, F1-Score of 83.0%, and an ROC-AUC of 0.93. Furthermore, SHAP analysis identified Grade Point Average (GPA), Attendance Rate, Previous Semester GPA, Financial Status, and Learning Management System (LMS) activity as the most influential factors affecting student dropout prediction. By combining high predictive performance with interpretable explanations, the proposed Early Warning System provides transparent decision support for lecturers and academic advisors, facilitating data-driven interventions aimed at improving student retention in higher education institutions.