<b>Understanding the Impact of Urban and Peri-Urban Park Environments on Violent Crime in Wuhan, China: A Machine Learning Approach with XGBoost-SHAP</b>

Violent crime poses a significant threat to the utilization of urban park amenities and access to their benefits. While some studies have utilized machine learning to explore the impact of park environments on crime, the underlying mechanisms remain underexplored due to the black-box nature of these models. Furthermore, prior research has predominantly focused on central urban parks, neglecting the critical role of peri-urban parks. This study employs the XGBoost-SHAP method to examine the varying effects of peri-urban and urban park environments on violent crime in Wuhan, China. We assessed four key variables: accessibility, built environment, land use, and socioeconomic factors, to understand their influence on crime rates. Our findings reveal that violent crime rates are higher in urban parks, although the park location alone does not significantly predict crime. Notably, both park accessibility and the built environment exhibit significant nonlinear effects on violent crime. Despite the potential for effective environmental design to mitigate crime, disadvantaged communities often experience poor safety conditions in their parks. These results underscore the complex interaction between park environments and violent crime, advocating for the application of machine learning interpretation to deepen our understanding of crime dynamics in parks.