Built Environment Factors and Interventions Supporting Children’s Active School Travel: Insights from Machine Learning

Active school travel (AST) is an important source of daily physical activity for children, yet its environmental correlates remain mixed and may be more complex than conventional linear models suggest. This study examines the associations of school-travel distance, built environment, socio-demographic attributes, and natural environment with children’s AST in Wuhan, China. Using 2020 travel survey data for children aged 6–15 and an XGBoost–SHAP framework, the study evaluates relative importance, nonlinear association patterns, and pairwise interactions among key variables. The results show that school-travel distance is the dominant correlate of AST, accounting for 44.07% of total model importance. Built-environment variables collectively contribute 36.24%, substantially exceeding socio-demographic attributes, while terrain also shows a non-negligible role. Several built-environment variables, including transit station density, road density, intersection density, land-use entropy, and POI density, display clear nonlinear patterns and directional shifts. Distance and age further condition the associations of other variables with AST, indicating substantial heterogeneity across distance bands and developmental stages. Overall, the findings suggest that AST should not be understood through average main effects alone, and that more age-sensitive, distance-sensitive, and context-specific planning strategies are needed to support child-friendly school travel.