An Integrated Machine Learning and AIGC Framework for Student Performance Prediction and Personalized Pedagogical Support in LowResource Higher Education: Evidence from Sierra Leone

Higher education institutions (HEIs) in Sierra Leone face persistent challenges in monitoring student academic performance and improving teaching quality, constrained by manual record-keeping, delayed feedback, and limited capacity for data-driven decision-making. This study proposes and validates an integrated software framework combining Machine Learning (ML) predictive analytics with Artificial Intelligence Generated Content (AIGC) for automated pedagogical support, tailored to low-resource environments. Three ML classifiers Logistic Regression, Random Forest, and Gradient Boosting were evaluated using the Open University Learning Analytics Dataset (OULAD; n = 6,519) as a simulation proxy. An AIGC module employs structured prompt engineering to transform ML outputs into context-sensitive instructional feedback. Gradient Boosting achieved the highest overall accuracy of 88.94% (weighted F1 = 0.89) across three risk categories. Binary pass/fail classification reached 93% accuracy. Assignment submission timing (avg_date_submitted) was the dominant predictor (importance score: 0.490). The AIGC module produced coherent, stakeholder-differentiated feedback. The proposed ML+AIGC framework demonstrates technical feasibility for early-warning and personalized pedagogical support in low-resource HEI contexts. Its lightweight, modular design offers a replicable blueprint for responsible AI adoption in Sub-Saharan African higher education and similar environments globally.