Explainable Optimized Machine Learning for Predicting Student Self-Regulated Learning

Abstract

This study examined the utility of an optimized, explainable machine learning (ML) model for predicting students’ self-regulated learning (SRL). Participants were 823 middle school students (52.6% girls) from randomly selected public schools who completed in-person measures of relevant social, motivational, and contextual factors. Results showed that a tree-based ML model combined with explainability techniques showed strong internal predictive performance and identified influential predictors of SRL. Task interest was the strongest predictor of SRL, followed by subjective well-being and perceived teacher support, with moderate contributions from responsibility for learning and smaller effects of social well-being. These findings demonstrate the value of explainable ML for identifying key contributors to SRL and translating predictive models into actionable educational insights.