Committee on Law and Justice Statistics

• Activity Number: 284 - So You Think You Can Predict Crime? Lessons Learned from the NIJ Spatiotemporal Crime Forecasting Competition
• Type: Invited
• Date/Time: Tuesday, July 31, 2018 : 8:30 AM to 10:20 AM
• Sponsor: Committee on Law and Justice Statistics
• Author(s): Joel Hunt* and Patryk Miziula* and George Mohler* and Tuanjie Tong* and Dylan Fitzpatrick*
• Companies: National Institute of Justice and deepsense.ai and IUPUI and Intuidex, Inc. and Carnegie Mellon University
• Keywords: Crime; Justice; National Institute of Justice; Spatiotemporal; Prediction; Machine learning

Abstract

This session will focus on the 2016-7 National Institute for Justice (NIJ) "Real-Time Crime Forecasting Challenge," which challenged contestants to predict police calls-for-service in Portland, Oregon for four crime categories over five time periods. The competition attracted submissions from an extremely diverse audience including students, university labs, small businesses, and dedicated crime prediction companies. This session will bring together the top winning contestants, NIJ administrators, as well as predictive policing experts to discuss critical topics in crime forecasting.

Spatiotemporal modeling and prediction is of critical importance to many relevant fields. This is especially true for urban environments where dynamics are shaped by complex geographic and demographic features. Crime forecasting presents a uniquely challenging and policy-relevant domain. There is keen interest in both the statistical and government communities for advanced crime prediction algorithms, and cities across the world are beginning to employ such systems to direct their policing activity.