Developing a Predictive Analytics Model for Early Identification of At-Risk Students in Nigerian Open and Distance Learning Institutions

This study develops and validates a predictive analytics model for early identification of at-risk students in Nigerian open and distance learning institutions, making an original contribution to learning analytics theory and practice in ODL contexts in Africa. Student dropout and underperformance in Nigerian ODL institutions represent major systemic challenges undermining access equity goals. Predictive analytics using machine learning techniques offers the potential to identify at-risk learners early enough for targeted intervention. A sequential model development design was employed using seven years of anonymised student data from the National Open University of Nigeria covering 15,000 student records. The dataset included demographic variables, prior academic performance, course engagement metrics, financial records, and study centre location. Feature engineering, exploratory data analysis, and machine learning model comparison across logistic regression, random forests, and gradient boosting algorithms were conducted using Python. The gradient boosting model achieved the highest predictive accuracy at 84 percent and AUC of 0.91 in distinguishing at-risk from non-at-risk students by the end of the first semester. Key predictors include early assessment submission patterns, login frequency in the first six weeks, and programme-type interaction effects. Validation on a held-out dataset confirms model robustness. The study presents an original validated predictive model and recommends NOUN establish a dedicated student analytics unit to operationalise early warning systems. Keywords: predictive analytics, at-risk students, open and distance learning, NOUN, machine learning