Influencing factors of urban safety perception based on the combination of multi-source data and machine learning: a case study of Nanchang City, China

Abstract

Abstract Research examines the influence of multisource urban data on residents’ perceptions of safety. Utilizing the SHAP machine learning model, the research conducts a comprehensive analysis of the nonlinear relationships and interactive effects between built environment factors and psychological perception factors on urban residents’ safety perceptions. Focusing on Nanchang City as a case study, the research integrates multidimensional data encompassing urban spatial environments, resident perceptions, and socioeconomic indicators. The findings highlight the critical role of perceived urban vitality and perceived wealth in shaping residents’ safety perceptions, addressing the insufficient consideration of individual psychological factors in previous research, this study innovatively incorporates psychological perception data, thereby extending traditional built environment theories. By employing nonlinear models to elucidate the influence mechanisms of different variables across spatial zones, it provides a scientific foundation for urban planning and safety governance. Additionally, selecting Nanchang, a representative medium-sized city characterized by historical and cultural heritage, as the sample addresses the previous research gap in such urban contexts. By establishing an evaluation framework for urban safety perception based on multi-source data, this study offers theoretical support and practical guidance for precision planning in medium-sized cities dominated by historical and cultural heritage. This contributes to advancing sustainable urban development and enhancing residents’ well-being.