Classification of Numeracy Achievement of Junior High School Educational Units Based on National Assessment Data using Random Forest

Abstract

This study classifies numeracy achievement in Indonesian junior high schools using 2023 National Assessment data from 11,399 schools. The Random Forest algorithm was applied because it is able to capture nonlinear relationships and complex interactions between heterogeneous predictors, while simultaneously reducing variance through bagging and out-of-bag validation techniques. Two models were developed, one without and one with literacy variables. The addition of literacy increased accuracy from 82.97% to 90.0% and increased the ROC-AUC value from 0.8986 to 0.9609. Based on Gini importance, literacy was the most influential predictor, followed by religiosity, learning experience, gender equality, and class size. Government policies need to integrate literacy and numeracy improvements within a unified curriculum framework and promote gender equality and contextual learning in schools. Furthermore, utilizing data-driven analysis from the National Assessment is crucial for guiding targeted interventions and equitable resource allocation for numeracy improvement.