With the right kind of data analytics, geologists can predict earthquakes. Turns out those same principles can be applied to predicting crime. That concept has been turned into an analytics tool that is now being used by more than 20 police departments in the U.S. and one in the United Kingdom.
Chief of R&D and co-founder of predictive policing company PredPol Jeffrey Brantingham explained the comparison to Government Technology. “Look at earthquakes. You have a complex terrain that has faults all over the place. … Big ones, small ones, and they tend to generate earthquakes at a known rate. … And when you have those earthquakes on those faults, they produce lots of aftershocks.”
Brantingham said that crime follows a similar pattern. “The mathematical architecture being used to study crime is closely related to what you can use to study earthquakes.”
Brantingham said that there are features in the environment that, by their nature, are prone to crime, much like faults are to earthquakes. “A shopping mall is a great example. It’s a standing crop of cars to be stolen or to be broken into. It’s not going anywhere. It’s like a big fault that produces earthquakes at a regular clip.”
Crimes also tend to produce “aftershocks,” as criminals tend to be creatures of habit and return to the same locations where they found previous success. “So you get these after-crimes that occur near in space and near in time to previous crimes.”
PredPol takes that information – what crime is being committed, when and where it happens – and applies mathematical algorithms, and uses it as the basis to forecast where crime will happen in the future. The concept grew out of using hot spots to track where crime is occurring (where you map crime over long intervals to see where it is concentrated to aid in the allocation of resources). But PredPol cuts down the reaction time to changes in crime patterns that might not be readily apparent by simply looking at hot spots marked on a map, thus allowing for a more real-time reaction to changes in the crime landscape.
But while the tool is about predicting crime, Brantingham also pointed out that this is not a profiling tool to identify who is committing crimes.
“We’re actually not saying anything about who, we are saying something about where and when crime is most likely to occur regardless of who may or may not be prone to commit those crimes,” he said.
Success in the Field
For the Santa Cruz, Calif., Police Department, statistics show an overall decline in crime in targeted areas. From 2011, when PredPol was implemented to 2012, burglaries declined 22 percent, robberies declined 27 percent, assaults were down 9 percent and arrests were up 56 percent while auto theft recovery improved by 22 percent. Available year-to-date totals through August 2013 show robberies down 4 percent and assaults down 11 percent, and though burglaries are up 7 percent for the year so far, they have seen a decrease in recent months.
Santa Cruz Deputy Police Chief Steve Clark noted that crime patterns are dynamic, but having PredPol allows for more real-time reaction to those changing patterns. He also suggested that the big drop from 2011 to 2012 accounted for the “low-hanging fruit” and expects the reductions to be less dramatic in years to come.
Clark noted that his department was an early partner of PredPol, helping validate the predictive modeling algorithms. He approached the company after reading an article about their research, hopeful it could help in Santa Cruz, particularly as the department was facing more calls for service and a decrease in “unobligated minutes” when officers can patrol while not on a call.
Faced with static resources, Clark said the department recognized that it needed to be more effective with its unobligated minutes to help keep communities safe and maintain quality-of-life standards.
The Tacoma, Wash., Police Department also is seeing a positive impact following the start of a year-long beta test that began March 1, 2013, though no statistics are yet available.
Speaking on the decision to use the tool, Assistant Chief Pete Cribbin noted that in 2005, the Tacoma Police Department began strategic planning activities, one of which was to find ways to leverage data for more effective policing.
“The police department was collecting enormous amounts of data but not effectively using [it] in crime prevention efforts,” Cribbin said. “The goal was to become a data-driven organization with real-time data available to all officers, investigators and command staff.” He noted that hot spot mapping and other tools were previously employed and PredPol has become an extension of those earlier efforts.
Tacoma Police Department Crime Analyst Megan Yerxa added that officers “have the ability to look at PredPol on their smartphones and laptops in real time and make adjustments as needed. It provides a much more focused way of policing that shifts the efforts beyond responding to crime to proactively preventing it by policing the probabilities for crime.” The department is currently using PredPol to target property crime, including residential burglary, commercial burglary, motor vehicle theft, vehicle prowls and robbery.
Cribbin anecdotally noted that because of adding the PredPol analytics tool, “we have had several successes of catching the offender ‘red-handed’ and making arrests.”
Santa Cruz’s Clark also pointed to some residual benefits in changing officer habits.
“Every officer out there knows a spot on their beat where they can go right now and arrest somebody,” Clark said. However, “it’s not really addressing some of the core issues that are relating to these other crimes that we are targeting in this. … It gives you a more purpose-based strategy as you go out there as an officer each day.”
And Clark was emphatic in stating that PredPol is a tool to enhance the work that officers do.
“This is not a substitute for your experience, your knowledge, your good old-fashioned gut sense of things,” he said. “This is here to help enhance that.”