Machine learning-driven analysis of academic performance determinants: Geographic, socio-demographic, and subject-specific influences in Somaliland's 2022–2023 national primary examinations

Abstract

This study examined factors influencing academic performance among primary school students in Somaliland. It utilizes data from 20,638 students who participated in the 2022–2023 national primary examination. The research employed a combination of machine learning algorithms and traditional regression methods to investigate subject-specific, socio-demographic, and geographic influences on achievement performance. The findings indicate that proficiency in mathematics and science are the strongest predictors of academic success. Performance exhibits significant variation by location, school type, and region. Urban students demonstrate superior performance compared to their rural counterparts, and private school students outperform those in public schools. Among the machine learning models evaluated, the Support Vector model proves the most effective for predicting outcomes, with an RMSE of 43.23 and MAE of 33.71. The regression model accounts for 77.9 % of the variance in performance, demonstrating the robustness of the predictors. This study highlights the inevitability for battered involvements to enhance STEM education and mitigate inequalities. It also underlines the potential of integrating machine learning with traditional analysis in resource-limited settings. These understandings can inform policymakers and educators in improving equity and quality in Somaliland's education system, thereby improving progress toward Sustainable Development Goal 4.1.