Predicting academic success: machine learning analysis of student, parental, and school efforts

Abstract

Understanding what predicts students’ educational outcomes is crucial to promoting quality education and implementing effective policies. This study proposes that the efforts of students, parents, and schools are interrelated and collectively contribute to determining academic achievements. Using data from the China Education Panel Survey conducted between 2013 and 2015, this study employs four widely used machine learning techniques, namely, Lasso, Random Forest, AdaBoost, and Support Vector Regression, which are effective for prediction tasks—to explore the predictive power of individual predictors and variable categories. The effort exerted by each group has varying impacts on academic exam results, with parents’ demanding requirements being the most significant individual predictor of academic performance; the category of school effort has a greater impact than parental and student effort when controlling for various social-origin-based characteristics; and significant gender differences among junior high students in China, with school effort exhibiting a greater impact on academic achievement for girls than for boys, and parental effort showing a greater impact for boys than for girls. This study advances the understanding of the role of effort as an independent factor in the learning process, theoretically and empirically. The findings have substantial implications for education policies aimed at enhancing school effort, emphasizing the need for gender-specific interventions to improve academic performance for all students.