<![CDATA[This study aims to compare the determinants of academic performance between Mathematics and Indonesian Language subjects using an Educational Data Mining approach. Utilizing the UCI Student Performance dataset, comprising 395 Mathematics and 649 Indonesian Language students, we identified 382 students who were enrolled in both subjects through identity-based merging. Random Forest models were optimized via Grid Search and evaluated using RMSE, MAE, and R², while SHAP (SHapley Additive exPlanations) was employed for model interpretability. The results indicate that the Random Forest model achieved higher predictive accuracy for Indonesian Language (R² = 0.85) compared to Mathematics (R² = 0.78). Feature importance and SHAP analyses revealed significant differential impacts: Mathematics performance was predominantly influenced by absences (r = -0.28), study time (r = 0.31), and weekend alcohol consumption (r = -0.22), whereas Indonesian Language performance was more strongly determined by maternal (r = 0.32) and paternal (r = 0.28) education levels. We conclude that educational intervention strategies should be tailored specifically to each subject rather than applying a generalized approach.]]>
Copyrights © 2026
