Carnegie Mellon University

photo of Pittsburgh police cruisers in downtown Pittsburgh

Pittsburgh Crime Hot Spot Project: Preventing Crime with Predictive Policing

Within its project portfolio that spans a broad range of topics, Metro21 completed in 2019 a project on predictive policing that has garnered attention given recent incidents. This research project was a partnership between CMU and the Pittsburgh Bureau of Police (PBP), the Pittsburgh Department of Innovation and Performance, and the Pittsburgh Department of Public Safety. The goal of the project was to reduce serious violent crime in Pittsburgh through prevention without increasing arrests by predicting locations — not individuals — at heightened risk of violent crime. Police patrol activity was then directed to those locations. The project was run city-wide and included all Pittsburgh neighborhoods. The project concluded in December 2019, and we are no longer sharing data with the police.

The prediction model specifically did not use racial, demographic or socioeconomic data. Nor did it use data on individual persons. The model only used crime offense data for crimes with victims and 911 calls for service. Using this information, CMU researchers identified chronic hot spots based on the number of serious crimes committed in an area using data from previous years. In each selected area, the researchers then identified temporary hot spots using an AI-based predictive model. The temporary hot spots identified high-crime locations that the police may have been unaware of, anywhere in the city. It provided police zone commanders with a new tool predicting hotspots to augment their existing process of directing patrol activity.

The tool was evaluated in the field, resulting in a 34% drop in serious violent crime in temporary hot spot locations, and a 24% drop in chronic hot spot locations during the duration of the experiment. There were only four arrests during 20,000 hotspot patrols. There was also no evidence of displacement of crime to other areas. Lastly, there was no cost to the PBP for this project, as it used discretionary time of existing uniformed patrol officers.

Metro21 is committed to research that is beneficial to society as a whole and to upholding the highest level of ethical standards. This project was evaluated by and complied with all university research protocols.

To read the full Metro21 statement on this project please visit 

To view the presentation from the June 26, 2020 Listening and Sharing Session on the project click hereTo view the recording of the session click here.




Pittsburgh Bureau of Police

Pittsburgh Department of Innovation and Performance


Wil Gorr (co-PI)
Emeritus Professor of Public Policy and Information Systems, Heinz College, Carnegie Mellon University

Daniel O’Neill
Associate Professor of Computer Science, Public Service, and Urban Analytics, New York University