Putting spatial crime patterns in their social contexts through a contextualized colocation analysis

Abstract

This study proposes a novel contextualized colocation analysis to examine spatial crime patterns within their social contexts. The sample includes all reported MCI crime incidents (i.e., assault, break and enter, robbery, auto theft, and theft over incidents) in the city of Toronto between 2014 and 2019 (n = 178,892). Following a stepwise clustering feature selection, we begin our analysis by regionalizing the city based on the relevant social context indicators through a ward-like hierarchical spatial clustering algorithm. Then, we use a modified colocation miner algorithm with a novel Validity Score (VS) to select significant citywide and regional crime colocation patterns. The results indicate that eating establishments, commercial parking lots, and retail food stores are the most frequent urban facilities in citywide and regional crime colocation patterns. We also note several peculiar crime colocation patterns across disadvantaged neighborhoods. Additionally, the proposed analysis selects the patterns that explain an average of 11% more crime events through the use of VS. Our study offers an alternative method for colocation analysis by effectively identifying crime-specific citywide and regional crime colocation patterns. It also prioritizes the identified colocation patterns by ranking them based on their significance.