A new geodemographic classification of the influence of neighbourhood characteristics on crime: The case of Leeds, UK

Abstract

Understanding spatial composition and characteristics of urban communities is an important factor determining the potential for social interactions between residents, greater chances for neighbourhood safety and policy planning. Until recently, neighbourhood classifications were traditionally census-based. This study presents a different approach in developing a community-based area classification for exploring the relationship between neighbourhood characteristics and crime through the integration of new data sources from social media with contextual variables from the traditional census. In this study, partition around medoids (PAM) clustering algorithm is used to partition 105 Leeds community areas into four distinct groups based on their crime rates. Silhouette index and adjusted rand index (ARI) were used to evaluate the validity of clusters internally and externally. The new typology developed, has revealed new insights demonstrating the relationship between social cohesion and crime rates in Leeds community areas. The new partitioned clusters also provide area-based specific information, on criminal activities that could help in policy planning for community building and better resource allocation towards reducing crime rates.