Noncognitive predictors of mathematics achievement: a machine learning approach using large-scale educational data

Mathematics performance is essential for educational and economic outcomes, and research emphasizes the role of noncognitive factors such as self-efficacy, motivation, and socioeconomic background. The aim of the study was to identify which noncognitive variables were the most effective predictors of students’ mathematics achievement. PISA 2022 background questionnaire data were analyzed using machine learning (ML) algorithms on a cloud-based high-performance computing system. The secondary data analysis covered 613,774 students across 81 countries and economies. The analysis identified mathematics self-efficacy (MATHEFF) as the most significant predictor among 130 noncognitive variables. Two of the top three predictors were associated with socioeconomic status (SES). The decision tree model achieved 71% accuracy in predicting mathematics achievement. These findings underscore the global relevance of self-efficacy and socioeconomic predictors and demonstrate the methodological value of ML in analyzing large-scale international educational data, providing evidence of practical and policy significance.