What Drives Math Anxiety? A Theory-Informed Machine Learning Approach to Identifying Key Predictors Among Adolescents

• In Event: AERA Poster Session Sunday 9:45 am
  In Poster Session: Social, Cultural, and Emotional Foundations of Student Learning and Motivation

Abstract

This study used machine learning to identify key predictors of math anxiety among 613,744 15-year-old students worldwide using data from PISA 2022. Guided by an integrative theoretical framework, 38 variables spanning student background, beliefs, teaching practices, and school climate were analyzed. Of the five models tested, XGBoost achieved the highest predictive accuracy. Math self-efficacy, emotional regulation, stress resistance, and supportive classroom environments emerged as the strongest predictors. Findings highlight that math anxiety is shaped by malleable personal and contextual factors, offering actionable insights for targeted interventions in education.

Author

• Wondimu Ahmed, University of Akron