Random Forest Hits 90.2% Accuracy Predicting Brazilian Student Performance
June 16, 2026
A Random Forest model trained on SAEB microdata — integrating student, teacher, school, and principal data — achieves 90.2% accuracy and 96.7% AUC classifying 9th-grade and high school proficiency. SHAP analysis identifies school-level socioeconomic status as the dominant predictor, outweighing individual student factors.
HOW THIS AFFECTS YOU
●
researcherThe multi-source feature integration and SHAP-based explainability approach is a replicable template for education policy ML studies in other national assessment datasets.