DEVELOPING AN INTEGRATED MACHINE LEARNING FRAMEWORK FOR PREDICTING DROPOUT RATES IN NIGERIAN PRIMARY SCHOOLS

Abstract

School dropout is a significant global challenge with severe socioeconomic consequences, particularly in developing nations like Nigeria. This study addresses this issue by developing a robust and interpretable machine learning (ML) framework for the early prediction of dropout risks in Nigerian primary schools. We propose an integrated pipeline that leverages multi-source attendance data from the Better Attendance and Mentoring Information System (BAMIS) across Northern Nigeria. The core of our framework is an ensemble Random Forest model, which is enhanced by a hybrid Synthetic Minority Oversampling Technique and Edited Nearest Neighbours (SMOTE+ENN) to mitigate class imbalance. A key contribution of this research is the use of attendance behaviour to engineer a proxy label for dropout status, as direct labels are unavailable. To ensure model transparency and practical utility for policymakers, we integrate SHapley Additive exPlanations (SHAP) to provide both global and local interpretability of the model's predictions. The framework aims to transform reactive dropout interventions into proactive, data-driven strategies. Our findings demonstrate the potential of this approach to provide actionable insights for school administrators and government agencies, enabling timely support for at-risk students and contributing to the achievement of Sustainable Development Goal 4, which aims to ensure quality education for all.