In responding to the U.S. Justice Department's National Institute of Justice's (NIJ's) challenge to develop a means for real-time crime forecasting, this report describes the methodology and findings presented by the Conduent Team (the Team), which deployed functionality from the Operational Analytics module within the Conduent Business Intelligence platform (CBI).
CBI is a configurable, flexible platform that provides a user-friendly interface for running machine-learning-based analytics. The project focused on crime data exported from a Record Management System (RMS). The Team further refined its predictive models with census tract data, which provides population, education, and OpenStreetMap data. These provide points of interest. The objective was to highlight fact-based, yet often unintuitive, actionable insights. The following analytical features were used: 1) clustering of law enforcement agencies with similar crime patterns, effectively peer-ranking predictive models; 2) benchmarking agencies based on relevant key performance indicators; 3) identifying the most important domain-relevant attributes, such as demographics and city characteristics, that influence crime patterns of the regions; and 4) hot-spot prediction to forecast the time and type of the next likely crime event. This report focuses on the technical details of this last feature in providing context to hot-spot prediction. The report concludes that the Team's efforts provide important insights into the practical aspects of detecting hot spots in real-time crime analytics, thus improving hot spot prediction. The team looks forward to workshops, demonstrations, and collaborations across companies and with NIJ subject-matter experts in deploying technology in new and creative ways to reduce the impact of crime. 3 figures
Report (Grant Sponsored)
Date Published: January 1, 2018