The NYPD has a new digital pattern-spotting partner to help connect crimes across precincts and find their suspect quicker.
The software, designed in-house over two years and dubbed Patternizr, automates traditional police leg-work through machine learning to find patterns in crimes.
“The existing process was a little bit time intensive,” Evan Levine, assistant commissioner of data analytics in the NYPD, said in a podcast, Resoundingly Human. “[Patternizr] gives the analyst a good head start in finding these patterns.”
Levine said crime patterns are a critical part of police work and this program hones investigators, saving time and money.
“Once you identify that a group of crimes are related, you can share evidence across these cases, you can focus investigative resources on them and solve the cases much quicker,” he said.
Levine, along with co-creator Alex Cholas-Wood, former director of analytics for the NYPD, first detailed the software last month in The Institute for Operations Research and the Management Sciences. They said the idea came from a New York University study on trend-spotting that didn’t turn out a workable program.
“We had some weakness in finding patterns that span those large geographic areas,” Cholas-Wood said in the podcast, noting the prior issue of connecting crimes details in one precinct to another.
The software, which is a group of machine learning algorithms from 10 years of data, scans through information on crimes, including distance, time, method of entry and type of force and spots patterns.
In about a third of the time, Patternizr was able to fully re-create old crime trends during testing. Eighty percent of the time it found at least some of the pattern.
The software does not capture race, gender or specific location of crimes.
“We felt that it was important to publish the specifics of the algorithm in the journal because we want to show that we are coming up with new ideas to improve public safety but we are doing it in a way that is responsible,” Levine said.
The New York Civil Liberties Union told the Associated Press it had not reviewed Patternizr, which rolled out in December 2016, but asked for the NYPD to be transparent about its use.