Online Exclusive

Software can help prevent crime one byte at a time

June 10--Handcuffs, sidearms and old-fashioned street smarts have long been tools used by law enforcement officers.

In the 21st century, they have added sophisticated computer analyses to track, predict and in some cases prevent crime.

The Knoxville Police Department and the Knox County Sheriff's Office have crime analysis units that sift through data from the thousands of crime reports officers submit during the course of a year. They look for crime patterns so they can identify hot spots for certain offenses and find links between crimes that otherwise might escape notice.

Neither agency has invested in the most promising -- and expensive -- computer software that can help predict crime, though both are investigating doing so. We urge them to find a way to add sophisticated analytical software to their tool chests.

Crime statistics have been used in the past as a way to measure the extent of a jurisdiction's crime problem and how past efforts affected the crime rate. Now, agencies use statistics on where and when crimes occur to assist in determining staffing levels for various shifts and patrol areas.

Predicting where and when crimes will happen in the future is the next frontier.

KPD is considering three firms to build a model that would help predict crimes, according to Deputy Chief Gary Holliday, who runs the agency's analysis unit along with civilian employee Jonne Crick. One of them, PredPol, was highlighted in a recent News Sentinel article on the promise and limitations of analytical crime fighting.

Founded in 2012, the company makes software that is the result of a collaboration between researchers and the Los Angeles Police Department. They found that criminal acts, like earthquakes, tend to fall along "fault lines" where certain types of crime are common.

Jeff Brantingham, a researcher at the University of California, Los Angeles and one of the founders of PredPol, told the News Sentinel that the software predicts criminal activity within 500-foot-by-500-foot boxes throughout the city.

No system is infallible, of course. Brantingham said violent crimes such as domestic violence and murder resist predictive analysis because they happen so rarely. And officers must use the information intelligently -- without abandoning the street smarts that are indispensable to their jobs.

KCSO's analysis unit, run by Capt. Bobby Hubbs, is looking at a software model developed at Temple University that is much more limited that PredPol but would come free, with some restrictions, because it was developed using grant funding.

Hubbs is skeptical of relying too much on software, noting that predicting human behavior is difficult and other methods are effective. For example, collaborating with probation and parole officers to track criminals being released back into the communities can be useful, he said.

While there are limitations to using predictive software, KPD and KCSO should continue to explore its use. Smarter policing through prevention helps make our communities safer.

Copyright 2014 - The Knoxville News-Sentinel, Tenn.