IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Predicting Crime Using Analytics and Big Data

A unique approach to crime analysis may allow police officers to predict illegal activity.

Police officers sometimes feel like they’re playing Whac-A-Mole — they work to decrease illegal activity in a high-crime area, only to have it pop up somewhere else. This is how Joel Caplan, an assistant professor in Rutgers University’s School of Criminal Justice, described the feelings of some of the police officers that he works with on crime research.

“They identify a hot spot, they suppress crime in that location, and they realize that it emerges in other places or comes back the moment they leave,” Caplan said. “They’re trying to figure out how to manage crime in a way that produces long-term positive outcomes.”

To solve that problem, Caplan partnered with two Rutgers colleagues, professors Eric Piza and Leslie Kennedy, to research risk terrain modeling. The approach takes an area and blends its history of crime with data on local behavioral and physical characteristics to create a map of locations with the greatest crime risk.

Whereas law enforcement’s traditional hot-spotting approach focuses mainly on the historical concentration of crime, risk terrain modeling examines the factors that contribute to this concentration and helps police officers predict where new hot spots could arise using data that’s often ignored or taken for granted.

“It paints a picture of those underlying features of the environment that are attractive for certain types of illegal behavior, and in doing so, we’re able to assign probabilities of crime occurring,” Caplan said.

The new approach will soon be put to the test. In February, the National Institute of Justice (NIJ) awarded Rutgers a two-year, $500,000 grant to conduct risk terrain modeling research in U.S. cities.

Caplan and his colleagues in the Rutgers Center on Public Security are working with police forces in Arlington, Texas; Chicago; Colorado Springs, Colo.; Glendale, Ariz.; Kansas City, Mo.; and Newark, N.J., to map and analyze local crime as part of a pilot project. The project’s collaborators hope the data will help officers suppress crime efficiently, and that law enforcement agencies in other jurisdictions will adopt risk terrain modeling once the technique is proven.

Caplan is confident that risk terrain modeling will support officers’ existing opinions and hunches about local crime. “It makes sense to them at this basic level of being able to articulate and standardize their own gut feelings and perspectives,” he said. “Information can be communicated in a consistent way among other officers in their department and across jurisdictions and provide useful information that might have otherwise been difficult to articulate or convey prior to this mapping technique.”

Demystifying Risk

Risk terrain modeling examines how the environment affects illegal activity. In December 2012, Rutgers researchers applied their approach to the crime data displayed on old maps of Irvington, N.J., to demonstrate the method. They merged Esri’s GIS software with police data and city maps to create new maps displaying the migration paths of neighborhood shootings in 2007 and 2008.

In one portion of their demonstration, the researchers analyzed maps displaying spatial relationships between shooting incidents in 2007 and nearby clusters of bars, clubs, liquor stores and fast food restaurants, which were designated as crime risk factors. All shootings occurred within blocks of those crime risk factors, so researchers concluded that the presence of these businesses caused higher crime rates.

In another set of maps, researchers displayed the migration paths of shooting incidents throughout 2007 and 2008. Those results suggested that criminals shift their illegal activities from one high-risk area to another over time.

The Rutgers researchers said that this modeling approach allowed them to objectively predict the ebb and flow of Irvington’s criminal activity.

“It allows you to not make any assumptions about what the underlying qualities of crime are,” Piza said. “The technique allows you to tangibly collect data about whatever features that you suspect may be contributing to the crime problem and then empirically test whether or not that’s actually the case.”

The researchers were careful to use software that was within the financial reach of the average police department. They gathered data using IBM’s SPSS predictive analytics software and Epi Info, statistical software developed by the Centers for Disease Control and Prevention, before they applied Esri GIS tools. Caplan said that other GIS tools could also handle the mapping functionality.

The Rutgers researchers provide a wealth of risk terrain modeling resources openly to taxpayers and law enforcement, regardless of whether they’re partners in a project. The Rutgers Center on Public Security website,, features videos and more than 5,000 downloads that educate readers and viewers about the modeling approach and its potential benefits.

Putting Theory Into Action

Over the next two years, the researchers hope to produce results from the pilot that will prove risk terrain modeling’s worth to other crime researchers and police officers around the nation.

“We’re not selling just a methodology. We’re selling something that’s much more complete, that’s theoretically driven, that has empirical tests involved and that creates a product that people can use,” Kennedy said. “I think that type of rigorous evaluation, self-analysis and peer review has really helped us a lot in terms of the credibility of the product.”

Over time, Rutgers’ School of Criminal Justice has built relationships with crime analysts and law enforcement officers. Some of those relationships led to partnerships for the current pilot project. For instance, Sgt. Jonas Baughman of the Kansas City Metro Patrol Division said his jurisdiction had a professional relationship with Rutgers for at least three years before agreeing to become one of the pilot sites. Baughman first contacted Caplan several years ago, after reading one of the professor’s articles on risk terrain modeling.

“Dr. Caplan reached out to me about his NIJ project, extending the invitation with some of the other law enforcement agencies to collaborate in the larger group project,” Baughman said.

The project will be in the planning phase for most of 2013, with deployment occurring the following year. Kansas City is still determining exactly how the city’s personnel will be working with Rutgers and the NIJ on the project details.

“Most likely it’s going to be a select few people who are going to actually do the data work, but it’s meant to be a patrol-centered project. I’m sure our robbery detectives will be involved at some point,” Baughman said. “It’s meant to be a very open project where it’s not just people crunching numbers and computers. It’s meant to be little bit of work in the office, but then we’re applying that work from the office to the field.”

More Research to Come

Caplan, Kennedy and Baughman didn’t specify how they will gauge the project’s success rate, but said their overall goal with the NIJ grant is to see how valuable risk terrain modeling will be in stopping crime.

“The second part of this grant is to assess how well the information produced from risk terrain modeling can be used to develop and implement risk-based interventions, and then to evaluate the outcomes in terms of how well crime is affected, suppressed or prevented in response to the risk terrain maps and information that’s produced,” Caplan said.

Ultimately they won’t know how valuable their data will be for a while, but Kennedy is confident that they’ll be pleased with the results.

“That’s something we’re waiting to see, but we have six sites, and they’re different-sized cities, and they’re different types of police forces,” he said. “We’re confident that we’re going to come up with convincing results from this whole process.”  


Miriam Jones is a former chief copy editor of Government Technology, Governing, Public CIO and Emergency Management magazines.