Effective enrolment forecasting and resource allocation are critical challenges for higher education institutions in developing countries, where financial constraints, infrastructural limitations, and volatile admission patterns significantly complicate strategic planning. This research investigated the application of machine learning techniques to improve student enrolment prediction accuracy and optimise institutional resource allocation at the University of Ibadan. The research adopted a quantitative, data-driven approach guided by the Cross-Industry Standard Process for Data Mining (CRISP-DM) framework. The dataset comprised of comprehensive institutional records from 2014 to 2024, including undergraduate admissions, enrolment figures, faculty staffing, departmental budgets, hostel allocations, and macroeconomic indicators. A hybrid modelling strategy benchmarked multiple supervised learning algorithms, including Random Forest, Support Vector Machines, Extreme Gradient Boosting (XGBoost), Logistic Regression, and Long Short-Term Memory (LSTM) networks, across classification and regression tasks. Rigorous feature engineering, multicollinearity assessment, and 5-fold crossvalidation ensured model robustness and generalisability. Results showed that ensemble methods, particularly Random Forest, delivered superior performance, achieving classification accuracy of approximately 95% and regression R² values close to 0.93. Feature importance and SHAP (SHapley Additive exPlanations) analysis revealed that internal institutional factors, specifically faculty staffing levels, departmental budget allocations, student–staff ratios, and admission selectivity, exert substantially stronger predictive influence than external macroeconomic variables. The models also identified structural inefficiencies in resource distribution and quantified enrolment sensitivity to capacity constraints and disruptions. Building on these predictive insights, an optimisation framework was developed to recommend data-driven resource allocation strategies aligned with forecasted enrolment patterns. An interactive dashboard was designed to deliver actionable forecasts and optimisation recommendations to university administrators. The research demonstrated that integrating predictive analytics with resource allocation modelling produces a reliable decision-support system capable of guiding faculty staffing, budget distribution, and infrastructure planning under conditions of uncertainty. The findings underscore the strategic value of data-driven governance in higher education and provide a scalable analytical blueprint for similarly constrained universities in developing contexts seeking to enhance operational efficiency, financial sustainability, and long-term academic planning.
Samuel Oyedokun (Tue,) studied this question.