The predictive role of sedentary behavior and physical activity on adolescent depressive symptoms: A machine learning approach

Objective: This study aims to investigate the predictive value of sedentary behavior and physical activity in adolescent depressive symptoms. Methods: A total of 2419 adolescent students (grades 7–12) from six administrative regions in China were surveyed. Measures included the Physical Activity Rating Scale for Children (PARS-3), a self-designed questionnaire assessing sedentary behavior among Chinese children and adolescents, and the Children's Depression Inventory (CDI). Machine learning models were trained and tested to predict depressive symptoms based on different types of sedentary behavior, physical activity, and other key variables. Results: The trained random forest model demonstrated high predictive accuracy (ACC = 90.52 %), with a precision of 92.01 %, recall of 87.95 %, and an F1 score of 0.90. Key predictors of depressive symptoms included sedentary behaviors such as multimedia learning, watching TV, classroom learning, and playing video games. Physical activity also emerged as a significant factor in predicting adolescent depressive symptoms. Conclusions: The machine learning-based predictive model exhibited strong performance, suggesting that sedentary behavior and physical activity data can effectively predict depression symptoms in Chinese adolescents.