Researchers from the University of Chicago have developed an algorithm they say can predict individual incidents of crime within 1,000 feet (roughly 333 metres) one week before they take place.
The team behind the research recently published their findings in a Nature Human Behaviour journal article.
Lead author Ishanu Chattopadhyay explained they created a digital twin of urban environments to make the predictions.
“If you feed it data from what happened in the past, it will tell you what’s going to happen in the future,” Chattopadhyay said.
The algorithm was trained using historical public data on violent and property crimes in Chicago.
Violent crimes included homicide, assault, and battery, while property crimes encompassed burglaries, thefts, and motor vehicle thefts.
The researchers focused on these crimes because they were more likely to be reported in urban areas with a high level of distrust and lack of cooperation with law enforcement.
They also excluded crimes which could be more susceptible to enforcement bias, such as drug crimes, traffic stops, and misdemeanour infractions.
The algorithm divides Chicago into spatial tiles, each 1,000 feet across.
It can predict that a specific crime will occur within a particular tile a week before its occurrence.
According to the researchers, the algorithm had a predictive accuracy of 90% — far greater than what has been achieved with similar models in the past.
“It’s not magical; there are limitations, but we validated it, and it works really well,” Chattopadhyay said.
It managed to perform with similar accuracy when using data from seven other cities in the US.
The algorithm also helped the researchers confirm bias in police responsiveness to crimes based on the area where a crime occurred.
The researchers cautioned against using their tool to direct law enforcement policies, but instead advised incorporating it within urban policies and policing strategies to address crime.
“You can use this as a simulation tool to see what happens if crime goes up in one area of the city, or there is increased enforcement in another area,” Chattopadhyay said.
“If you apply all these different variables, you can see how the systems evolve in response.”