Predictive Analytics for Student Dropout Risk in Public Secondary Schools in Niger State

Early identification of students at risk of dropping out of secondary school allows targeted support that can significantly improve educational attainment in Niger State, where secondary completion rates remain below national targets. This study developed predictive models for identifying dropout risk among Junior Secondary School students across 25 government secondary schools in Minna, Bida, and Suleja Local Government Areas, Niger State, North Central Nigeria. Longitudinal student records covering attendance rates, academic performance scores, distance to school, household socioeconomic proxies, and teacher interaction logs were provided by the Niger State Universal Basic Education Board. Three models, namely logistic regression, random forest, and gradient boosting, were trained and evaluated. The gradient boosting model achieved the highest AUC of 0.86 and F1-score of 0.78 for dropout prediction. Attendance rate below 70 percent in a term was the most predictive feature, followed by failure in two or more core subjects and distance exceeding eight kilometres. Female students in rural LGAs showed the highest dropout risk scores. Seasonal patterns indicated elevated dropout probability in the final months of each academic year. Model explanations generated using SHAP values identified individualised risk factor profiles for each student. Recommendations include deploying the model as a teacher alert system through the state's school management information platform, establishing a community mobilisation programme for at-risk female students, and using the findings to inform conditional cash transfer programme targeting criteria in Niger State.

Keywords: student dropout prediction, gradient boosting, SHAP, Niger State, educational analytics