Exploring Predictors of Adolescents’ Creative Thinking Ability: using AutoML and SHAP Analysis

This study aimed to explore the predictors of creative thinking skills among Korean adolescents. To achieve this, the creative thinking responses of 6,117 high school students aged 15 years who participated in the Programme for International Student Assessment (PISA) 2022 were analyzed. Automated Machine Learning (AutoML) was applied to compare the performance of various machine learning algorithms, and key predictors were identified based on the best-performing model. In addition, SHAP (SHapley Additive exPlanations) values were used to assess the importance of the predictors. First, among various regression models applied to predict creative thinking ability, LightGBM demonstrated the highest predictive performance. Second, based on SHAP importance scores, the most influential predictor was the “number of mathematics classes per week”, followed by “familiarity with mathematical concepts”, “ICT use for topics beyond school subjects”, “mathematical self-efficacy: formal and applied mathematics”, and “expected academic achievement.” Finally, based on the results of this study, suggestions to promote creative thinking are discussed, and implications for future research are suggested.