The influence of neighborhood environments on children's travel mode choices: An XGBoost/SHAP model analysis of Shuangliu District, Chengdu, China

In contemporary urban planning, understanding how the built environment shapes children's travel mode choices is fundamental to the development of truly child-friendly communities. Previous studies rarely investigated complex interactions between multiple environmental features. Drawing on 3227 travel events undertaken by 1127 children across 118 neighborhoods in Chengdu, China, the study uses Extreme Gradient Boosting and Shapley Additive explanations model to explore the non-linear relationship between the neighborhood-built environment and children's travel mode choices. Three-dimensional partial dependence analysis reveals that walking propensity peaks only when Land use mix, green space index, and blue space index collectively meet optimal ranges. Walking is most likely when green space indices fall within the range 0.3–0.4 or 0.7–0.8, blue space indices range from −0.7 to −0.65, and where land use mix exceeds 0.7. These findings offer clear guidance for planners, identifying integrating mixed-use zoning with green corridors and minor water elements as effective strategies for promoting the active travel of children. By moving beyond linear, single component approaches, this study provides a nuanced framework for designing supportive environments that enhance children's independence, health, and well-being.