Where Will Criminals Strike Next?

​PREDICTIVE ANALYTICS use past events and data to predict what will happen in the future. This may conjure visions of Tom Cruise in the movie Minority Report, where premonitions allowed “criminals” to be arrested before actually committing their crimes, but the use of predictive analytics is no longer reserved for science fiction.

Numerous police departments are employing predictive analytics software from places such as IBM subsidiary SPSS. One department is the Memphis (Tennessee) Police Department (MPD), which integrated the software into a crime-fighting system called Blue CRUSH (Crime Reduction Utilizing Statistical History). Those connected with the department credit Blue CRUSH for dramatic decreases in crime.

In the Blue CRUSH system, which was developed with the help of Richard Janikowski, director of the University of Memphis’s Center for Community Criminology and Research, officers gather data on problem areas in the city and submit the data to the SPSS software, which then calculates the frequency and timing of crimes. What that software essentially does to the information, says SPSS’s Colin Shearer, is formulate the probabilities that crimes will occur in certain places at certain times. That helps the police department determine a risk level associated with the particular area. The information is used by the department in determining resource allocation to those areas.

Although the plans for assigning officers to areas based on the information are drawn up weekly, MPD Crime Analysis Manager John F. Williams says he likes to keep the effort as “real time” as possible. If there are crimes popping up in other areas or a pattern of crime occurs in a short period of time, changes are made accordingly.

The effort is about more than data analysis, says Janikowski; it also involves introducing a data-driven culture to the department. He says it even provides the opportunity to forecast future problems, such as what the effect of a cluster of vacant properties might be on crime.

Violent crime has dropped about 28 percent since 2006 when the MPD began using the analytics, says Williams. A recent example of Blue CRUSH’s effectiveness was in the use of the detailed heat mapping feature to address major crime issues in an area where the department ended up arresting around 50 drug dealers. And recently, the department used Blue CRUSH to determine where to place surveillance cameras.

Shearer says police departments and organizations use the software and information derived from it differently. For example, a United Kingdom police department uses survey data from citizens to find out which crimes people are concerned about; this information allows the department to determine whether they are addressing local needs.

In Florida, the juvenile justice system is attempting to use predictive analytics to determine whether juvenile criminals are likely to reoffend, and it is working that data into sentencing and assistance programs.

Predictive analytics programs also have detractors. Many worry they can encourage profiling or even punishment for crimes that haven’t yet been committed. The Florida program mentioned by Shearer has particularly attracted criticism recently.

Shearer says those characterizations are off base. “What we’re trying to do in all areas of the world is trying to help people make better decisions, have better information for them to base their decisions on. So, we’re not saying let our system tell people ‘this person is about to cause an offense—go and do something.’ Rather, what our system is doing is giving more input,” says Shearer, adding that the information can help officers make decisions, but won’t be the final say.

An officer takes predictive analytics into account along with his experience. “He’s not trying to hand over control to some all-knowing machine,” says Shearer.

Janikowski says that the department went through extensive training and many community meetings regarding Blue CRUSH. “There’s a big difference between being proactive and harassing or profiling. And you need to be emphasizing that.” He adds that it’s critical to have direct communication with the community.

The next area Janikowski would like to see Blue CRUSH expand into is text analysis. He would like to see a way to put the information in incident and arrest records towards the data analysis and predictive analytics.