Ecological Network Analysis: Utilizing Machine Learning to Unravel the Effects of Multilevel Pathways of Moderate⁃to⁃Vigorous Physical Activity Facilitators Among School Children.

Abstract

The objective of the present study was to ascertain whether the association between moderate-to-vigorous intensity physical activity (MVPA) levels and individual, interpersonal, organizational, and environmental factors among school children is influenced by their attitudes toward emerging sports participants (ESP). To this end, machine learning (ML) was employed to analyze the data. This cross-sectional study, involved 655 child-parent pairs in Changsha City to assess children's MVPA. Data were collected via self-administered questionnaires, evaluating MVPA levels and attitudes from children and caregivers. Various statistical models, including random forest and LASSO regression, were utilized for analysis. The study revealed that boys engaged in more MVPA than girls. Most participants liked ESP, with significant teacher support noted. Random forest and LASSO regression models identified key factors influencing MVPA, with notable variability among non-achievers. The gradient boosting machine and K-nearest neighbors models demonstrated similar predictive performance. The final model, comprising 37 parameters, indicated significant relationships between variables, particularly highlighting the importance of school offerings ESP and living near sports field. This study concludes that offering ESP in schools, along with positive modeling and encouragement from caregivers and peers, effectively enhances children's participation in MVPA. Living near sports field also positively impacts MVPA levels.