Nonlinear, Interaction, and Spatial Effects of Travel Time to Urban Services on Urban Vitality: XGBoost with SHAP Value

Abstract Urban vitality has been identified as a crucial factor in creating sustainable and livable cities. Consequently, understanding the relationship between access to urban services and urban vitality is important for effective urban planning and design. Many previous studies have relied on simplified distance measures and linear models, overlooking temporal variations by travel mode and their nonlinear, interaction, and spatial effects on urban vitality. This study addresses these gaps by examining how travel times to urban services by walking, driving, and public transit affect urban vitality in Seoul; moreover, it identifies nonlinear, interaction, and spatial effects of access time via travel mode on urban vitality. Our eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanations (SHAP) models confirmed the following findings: (1) shorter travel times by walking and public transit to neighborhood hospitals, metro stations, and large retail stores were primary contributors to urban vitality; (2) travel time by mode of transport had nonlinear and interaction effects with density and access to public transit on urban vitality; and (3) access time by mode of transport had spatially varying effects on urban vitality. These results highlight the importance of adopting flexible, context-sensitive accessibility strategies that cater to the diverse mobility needs and local contexts, contributing to more equitable and vibrant cities.