AI-Driven Predictive Analytics for Evidence-Based Education Policy: Evaluating Age and Sex Effects on Student Performance in Nigerian Higher Education

Abstract

The rapid expansion of higher education systems globally has intensified the need for data-driven decision-making in educational policy. This study develops and evaluates an AI-driven data analytic model to inform evidence-based policy formulation in Nigerian higher education, specifically examining whether student age at university admission significantly affects academic performance. Using a quantitative research design with descriptive, inferential, and predictive analytics, we analyzed administrative records of 3,897 undergraduate students from Nigerian universities. Descriptive analysis revealed that students aged 19–25 years achieved the highest mean CGPA (3.01–3.08), while students aged 26+ years recorded the lowest (2.38). Female students consistently outperformed male students across most age categories (mean CGPA: 3.06 vs. 2.89). One-way ANOVA confirmed significant differences in CGPA across age categories (F(3,3871) = 9.306, p<0.00001). However, Pearson correlation analysis revealed no meaningful linear relationship between age and CGPA (r=0.012, p = 0.3785). Two-way ANOVA demonstrated a significant interaction effect between sex and age category (F(1,8) = 7.716, p = 0.024). Predictive modeling using linear regression (R2 = 0.0088, RMSE = 0.957) and random forest (R2=0.0154, RMSE=0.957) indicated that age and sex alone explain less than 2% of CGPA variance. These findings suggest that while age category differences exist statistically, age alone is not a meaningful predictor of academic success. The findings support the integration of AI-driven analytics into higher education policy systems to enhance evidence-based decision-making and targeted student support interventions.