Scientists from the University of Chicago have developed a new algorithm that can predict crimes a week before they happen. Forecasts are 90% accurate, and the location of the event is determined with an accuracy of up to 300 m
For predictions, the algorithm studies patterns in public data on violent and property crimes. Research was published in the journal Nature.
The effectiveness of the tool was tested on data about the city of Chicago. The audit concerned two categories of crime: violent (murder, assault, and beating) and property (burglary, robbery, and car theft).
Such data were most frequently reported to the police, even in areas with a historical distrust of law enforcement. Also, they are less dependent on the bias of law enforcement officers.
The new model examines crimes in terms of time and space coordinates of individual events and reveals patterns for predicting the next ones. It divides the city into sectors about 300 meters across and predicts crime in them.
“We created a digital twin of urban environments. If you feed it data from what happened in the past, it will tell you what’s going to happen in the future” says the lead author of the study.
He notes that the algorithm should be used as a complementary tool to city policies and police strategies to combat crime.
“Now 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. If you apply all these different variables, you can see how the systems evolve in response,” says the scientist.
The tool also revealed police bias in crime in different areas of the city. When the crime rate in more affluent areas increased, it led to more arrests. But this did not happen in poorer areas, which indicates an imbalance in the response of the police.