AI predicts crimes with 90 percent accuracy

An AI model has managed to correctly predict crimes in eight US cities a week before they occur with 90 percent accuracy, reports Ishanu Chattopadhyay of the University of Chicago, who and his team created a virtual "twin city" surveillance model using crime data in Chicago created from 2014 to late 2016.

The model, which has produced similar results in seven other cities, focuses on the types of crimes committed and where they were committed. Ninety percent of the predictions were accurate for geographic areas up to two-quarters in size, delineated by intersecting streets.

 Image source: Pixabay; Which

Chattopadhyay, who has his dates and his algorithm has made public hopes that they will be used for action at the state and national level and not as a direct tool for the police. In other words, he doesn't want police officers simply being sent to places where crime is predicted, but rather community-based action and proactive policies leading to crime reduction.