Social scientists from the University of Chicago have developed an algorithm that forecasts crimes in urban areas one week ahead of time with a 90% rate of accuracy.

To accomplish this, the algorithm divides a target city into 1,000 sq ft tiles and examines the historical data on violent and property crimes from those regions to make predictions about future incidents, according to the team.

Unlike similar and controversial algorithms designed for the same task, the University of Chicago team suggests that their model is different because it doesn’t examine crime as emerging from hotspots and spreading to other areas, but rather analyzes previous crime reports among many other factors.

In addition to successfully predicting crime likelihood in Chicago where the algorithm achieved a 90% rate of accuracy, the social scientists also tested the algorithm on eight other U.S. cities including Los Angeles, California; Atlanta, Georgia; and Philadelphia, Pennsylvania, where the algorithm achieved similar results.

The crime-forecasting algorithm is detailed in the article, Event-level prediction of urban crime reveals a signature of enforcement bias in US cities, which appears in the journal Nature Human Behavior.