*WINNER* Using Graph-based Knowledge Discovery to Detect Anomalous Patterns in Crime Data

Abstract

The naive approach to law enforcement is purely reaction-based. However, more effective preemption of criminal acts is made possible by analyzing datasets comprised of previous incidents; attributes such as time, location, and other specifics of the event serve as crucial details for predicting reoccurrence. Graphs are an appropriate approach for representing these elements, as they can provide structure and emphasize relationships that other techniques may fail to. In addition, graphs can highlight an underlying hierarchy between subjects if one is present. In particular, the Tucson Police Department has made their recent crime archives publicly available. Data from 2018 until the present is readily downloadable for modification, cleaning, and analysis. This dataset will allow data scientists to supplement law enforcement with predictive tools and a clearer picture of the threats they face. This, in turn, will help the TPD allocate resources where they are needed most.