Predicting Primary School Student Dropout Risk: A Machine Learning Framework for Early Intervention

Abstract

Student dropout in primary education is a critical global challenge with significant long-term societal and individual consequences. Early identification of at-risk students is a crucial first step towards implementing effective intervention strategies. This paper presents a machine learning framework for predicting student dropout risk by leveraging historical academic, attendance, and demographic data extracted from a primary school system. We formulate the problem as a binary classification task and evaluate multiple algorithms, including Logistic Regression, Random Forest, and Gradient Boosting, to identify the most effective predictor. To address the inherent class imbalance, we employ Synthetic Minority Over-sampling Technique (SMOTE). Our results, validated via stratified 5-fold cross-validation, indicate that the Random Forest model achieved the highest performance, with a recall of 0.91 ± 0.03, ensuring that 91% of truly at-risk students were correctly identified. Furthermore, we use SHAP (SHapley Additive exPlanations) values to provide interpretable insights into the model’s predictions, revealing that attendance rate, academic performance trends, and socio-economic proxies are the most salient features. This work demonstrates the potential of machine learning as a powerful decision-support tool for educators, enabling timely and data-driven interventions to improve student retention and completion rates.