The association of educational policy learning motivation and socio educational factors with student numeracy using structural equation modeling and explainable machine learning in Riau Province Indonesia

This study examines the relationships among educational policy, motivation, and socio-educational factors associated with students’ numeracy achievement by integrating Structural Equation Modeling (SEM) and explainable Machine Learning (ML) approaches. Using data from junior high school students in Riau Province, Indonesia, this study explores the direct and indirect associations among school policy, learning environment, teacher quality, parental support, learning style, and learning motivation with numeracy outcomes. The SEM results indicate that school policy is not directly associated with numeracy achievement; rather, it is indirectly related through pathways involving the learning environment and teacher quality, which are further associated with parental support and students’ learning motivation. Teacher quality shows both direct and indirect associations with numeracy achievement, while parental support emerges as a consistent predictor of learning style, learning motivation, and numeracy performance. Learning style is associated with both learning motivation and numeracy achievement, and learning motivation is identified as the strongest predictor of numeracy outcomes. To complement SEM, explainable ML models were employed to evaluate individual-level prediction and feature importance. The results indicate that learning motivation, parental support, and learning style are the most important predictors of numeracy achievement, while school policy and learning environment show lower direct predictive importance. Among the evaluated machine learning algorithms, the best-performing model (Support Vector Machine) explained 27.8% of the variance in numeracy achievement (R² = 0.278), indicating that a substantial proportion of the variability is attributable to other factors not included in the present model, such as students’ socioeconomic status, prior mathematical knowledge, and other contextual influences. Overall, the findings highlight the importance of integrated and context-sensitive educational strategies that emphasize student motivation, parental involvement, and teacher quality.