Fighting crime with data
predictive policing projects aim to do the seemingly impossible: lower crime without putting more officers on the streets
Police departments around the world are harnessing data to predict and prevent crime. Last year, the police department in Oxford, Alabama, USA launched a project that officials believe could slash the local crime rate. The cost? Just US$12,000.
The one-year pilot project makes use of predictive policing, employing data on previous crimes to predict what offenses are likely to happen—both when and where. As police departments of different sizes and locations are tasked with reducing crime in the face of shrinking resources, predictive policing is seen by some as an effective strategy to fight crime without adding officers. Projects to implement computer models that forecast crime are popping up everywhere from sprawling metropolises like Los Angeles, USA to small boroughs like Trafford, England.
Executing the project in Oxford involved putting a decade's worth of crime reports into a software program and training staff to use it. The software divides the city into 500-foot-by-500-foot (152-meter-by-152-meter) squares and predicts which squares are most likely to see an uptick in criminal activity based on the hour. Patrols start their shifts by consulting the software, and have been trained to frequent those predicted hot spots up to 10 percent more than other areas.
The pilot project is slated to finish in April, when analysts will sift through data to determine whether the initiative delivered intended business results and should be rolled out across the department.
“We're looking forward to seeing what else it can do,” project sponsor and Police Chief Bill Partridge told Alabama. After Kansas City, Missouri, USA implemented a predictive policing program, murders dropped 20 percent in 2014 from the previous year.
The introduction of predictive policing in Kansas City, Missouri, USA was followed by a reduction in murders.
TOP PHOTO BY GARVEY SCOTT/KANSAS CITY STAR/MCT/GETTY IMAGES
After Kansas City, Missouri, USA implemented a predictive policing program, murders dropped 20% in 2014 from the previous year.
The overall improvement in identifying areas likely to see crime by using predictive policing typically hovers between 5 and 20 percent better than regular best-practice crime-forecasting methods, according to researchers at the Rand Corp. Some police departments have reported crime reduction as high as one-third, particularly for property crimes like car theft.
“Some reports show that predictive policing projects can help police departments capture two or three times as much criminal activity as the crime analysts,” says John Hollywood, senior operations researcher, Rand Corp., Washington, D.C., USA. But that assumes project leaders are ensuring every dimension of the project—from stakeholder buy-in to training—is handled well and that what they did previously to find hot spots was far from ideal.
“Some reports show that predictive policing projects can help police departments capture two or three times as much criminal activity as the crime analysts.”
—John Hollywood, Rand Corp., Washington, D.C., USA
Any reduction in crime often comes with savings. St. Louis County in Missouri, USA plans to complete a US$45,000 project to implement a predictive policing program before mid-2016. “If you look at the cost of crime, if you prevent a handful of aggravated assaults, this recoups its cost pretty quickly,” St. Louis County Sgt. Colby Dolly, crime analysis unit supervisor, told the St. Louis Post-Dispatch.
Data analysis projects may be held up as a salve for ailing police departments, but they're no cure-all. The English city of Kent launched a crime-prediction software project in April 2013, yet saw crime spike by 11 percent over roughly the next 12 months.
Part of the problem is that innovative solutions such as predictive analytics and data mining are being grafted onto sorely outdated legacy systems. British police waste an estimated £221 million a year because they can't access computer systems and records outside of the office, according to The Economist. Uniform adoption represents another hurdle to maximizing intended project benefits, according to the Rand Corp.
“An agile approach naturally lends itself to implementing a tool that officers are going to be using daily and ongoing.”
—Jeremy Heffner, Azavea, Philadelphia, Pennsylvania, USA
British police waste an estimated £221 million a year because they can't access computer systems and records outside of the office.
Source: The Economist
Stakeholder buy-in can be an issue as well, but pilot projects and iterations can be ways to effectively demonstrate benefits, says Jeremy Heffner, a senior data scientist and product manager at Azavea in Philadelphia, Pennsylvania, USA. His company develops predictive policing software.
“The project sponsors might think they know exactly how they want it and think the requirements are nailed down, but then the team uses it for a couple of days or weeks and they want to revise it,” he says. “An agile approach naturally lends itself to implementing a tool that officers are going to be using daily and ongoing.” His team works in two-week scrums to refine software features.
“This area is quite new,” Mr. Heffner says of predictive policing projects. “I don't think anyone has all of the answers yet about the best way to implement these tools.” In the meantime, project practitioners would be wise to borrow a page from the playbook of successful projects and include both time for training and systems for ongoing metrics. —Kate Rockwood
PM NETWORK MARCH 2016 WWW.PMI.ORG
MARCH 2016 PM NETWORK