26 February, 2015
LAPD Fighting Crime Before It Happens
The LAPD Foothill division is attempting to predict and prevent crime before it even happens. How do they do this? Big data, of course.
UCLA Professor Jeff Brantingham and his team partnered with the LAPD to examine 13 million crimes recorded over 80 years in order to predict where crimes would occur in the future. By piecing together patterns of human behavior within the LAPD crime data, they believed they could determine which areas to more heavily patrol with police officers and when.
To do this, Brantingham collaborated with Professor George Mohler from Santa Clara University, who is an expert in forming pattern systems. They began by using a mathematical model that was already being used in California to predict earthquake aftershocks.
Just as there is a high probability that aftershocks will follow earthquakes nearby in both space and time, there is an elevated risk for crime, which travels to neighboring regions after one crime has already occurred. In other words, Mohler and Brantingham were studying the “aftershocks of crime.”
By incorporating the 13 million arrests over 80 years into this model, the LAPD is able to use this algorithm to begin predicting the future of crime in the area. To put this to test, patrol cars were given “mission maps,” which showed boxes of 500 square feet that predict where crimes were most likely to occur during their 12-hour watch.
The early results from this test already show there has been a 12% decrease in property crime and a 26% decrease in burglary in the Foothill precinct alone. As methods develop and the prediction model is updated in real time, LAPD expects the data will become even more accurate and thus useful.