Pedestrians are widely recognized as the most vulnerable group among road users and are more likely to be involved in severe traffic accidents. Therefore, creating a safer environment for pedestrians is crucial for protecting residents and promoting high-quality urban development. Traditionally, the factors influencing pedestrian traffic accidents have been analyzed from a micro perspective, focusing primarily on intersections and crosswalks. However, the effects of medium- and macro-level built environment factors, such as urban spatial structure, land use, and density, on pedestrian traffic accidents have not been comprehensively and systematically addressed. Recently, scholars in the fields of urban planning and geography have begun to explore the relationship between the built environment and pedestrian traffic accidents, often employing statistical models such as negative binomial regression and logistic regression. In recent years, machine learning methods have been increasingly applied to the studies of walking behavior. These studies have revealed a nonlinear relationship between the built environment and walking activities, and since pedestrian traffic safety is closely related to walking behavior, it is likely that similar nonlinear and threshold effects exist between built environment factors and pedestrian traffic safety. Investigating these nonlinear relationships and threshold effects challenges the traditional assumption of a linear connection between the built environment and pedestrian traffic accidents. It also provides valuable insights into the local characteristics of how built environment variables influence pedestrian traffic accidents, particularly from the perspective of marginal effects. In practical applications, empirical findings on nonlinear and threshold effects, especially regarding the relative importance and threshold ranges of various built environment elements, can offer more refined strategies and effective solutions for traffic management and urban planning.Yuzhong District in Chongqing, located at the confluence of the Yangtze River and Jialing River, covers an area of approximately 23.24 square kilometers. With its dense urban road network and compact land use, Yuzhong District has developed a unique and complex transportation system that integrates both underground and surface transportation. However, the increasing number of vehicles and relatively underdeveloped pedestrian infrastructure have exacerbated conflicts between pedestrians and vehicles. For these reasons, Yuzhong District serves as an ideal and representative reseach area. This study collects spatial big data from multiple sources, including pedestrian traffic accident data, road network data, land use data, mobile phone signaling data, and Points of Interest (POI) data. Using the Gradient Boosting Decision Tree (GBDT) model,the study explores the nonlinear relationships and threshold effects of built environment factors across five dimensions: road facilities, land use, proximity to facilities, spatial structure, and socioeconomic factors, on the frequency of pedestrian traffic accidents.The findings of the study are as follows: First, there are significant differences in the degree to which built environment factors influence pedestrian traffic accidents. Spatial structure, with the highest relative importance, has the greatest impact on pedestrian traffic accidents, followed by population density, road network density, and land use mix. Second, the relationship between each built environment factor and pedestrian traffic accidents exhibits nonlinear and threshold effects. For example, pedestrian overpasses contribute more significantly to pedestrian safety than zebra crossings. When the number of pedestrian overpasses exceeds four per square kilometer, pedestrian traffic accidents can be effectively reduced. Similarly, in areas where the land use mix exceeds 0.37, traffic management measures such as speed limits should be implemented to mitigate pedestrian traffic accidents.