Adopting data interpretation on mining fine-grained near-repeat patterns in crimes

Abstract

The near-repeat effect is a well-known phenomenon in crime analysis. The classic research methods focus on two aspects. One is the geographical factor, which indicates the influence of a certain crime risk on other similar crime incidents in nearby places. The other is the social network, which demonstrates the contacts of the offenders and explain ”near” as degrees instead of geographic distances. In our work, these coarse-grained patterns discovering methods are summarized as bundled-clues techniques. In this paper, we propose a knotted-clues method. Adopting a data science perspective, we make use of a data interpretative technology and discover that the near-repeat effect is not always so near in geographic or network structure. With this approach, we analyze the near-repeat patterns in all districts of the dataset, as well as in different crime types. Using open source data from Crimes in Chicago provided by Chicago Police Department, we find interesting relationships and patterns with our mining method, which have a positive effect on police deployment and decision making.