A Machine Learning and Explainable Artificial Intelligence Approach to Student Dropout Prediction Using Multidimensional Educational Data

Student dropout prediction is a critical task for educational institutions seeking to enhance academic performance and reduce attrition. This study presents a machine learning-based framework that integrates comprehensive preprocessing, baseline model evaluation, and advanced ensemble learning for accurate dropout prediction. Six classical classifiers Logistic Regression, Decision Tree, Random Forest, Support Vector Machine, K-Nearest Neighbors, and Naïve Bayes are assessed to establish baseline performance. To improve predictive effectiveness, a Light Gradient-Boosting Machine model is employed and further enhanced through a hybrid stacked ensemble combining Random Forest, Light Gradient-Boosting Machine, and Support Vector Machine, with Logistic Regression as the meta-learner. The system extends beyond binary classification by introducing a three-level risk categorization (low, medium, high), enabling more focused interventions. Explainable AI techniques, specifically SHapley Additive exPlanations and Local Interpretable Model Agnostic Explanation, are incorporated to provide transparent global and local factor interpretations. The framework is supported by an interactive dashboard, demonstrating strong predictive accuracy, interpretability, and practical applicability for early identification of at-risk students.