Exploring Urban Crime Through the Lens of an Adaptive Region Partitioning Technique and Built Environment Features

Abstract

In the realm of urban studies, the choice of regional partitioning schemes can significantly influence analytical outcomes, thereby affecting the precision and dependability of research findings. This paper proposes a new region partitioning method with multi-density distribution, namely the Adaptive Region Partitioning (ARP) method, and explores the impact of built environment features under different region partitioning methods on crime analysis. Compared with the traditional partitioning method, the research unit under the ARP method exhibits improved intra-regional homogeneity and inter-regional heterogeneity characteristics, more accurately maps the spatial distribution characteristics of the urban data, effectively mitigates the risk of multicollinearity, improves the accuracy of the regression model, and provides a new and reliable regional partitioning scheme for crime research. Finally, by applying the SHAP (SHapley Additive exPlanations) method, this study further reveals the extent to which different built environment features influence crime, providing important data support for urban planning and the formulation of crime prevention strategies.

Keywords

• Urban crime
• built environment
• region partitioning
• multi-density data
• SHAP