This algorithm can predict crime week before it happens with 90% accuracy: study

Algorithm studies a city's criminal data and arrest history, recognising the patterns in time and locations

By
Web Desk
A robotic arm collects a pre-packaged dish from cold storage, done according to the diner’s order at Haidilaos new artificial intelligence hotpot restaurant in Beijing, China, on November 14, 2018. — Reuters
A robotic arm collects a pre-packaged dish from cold storage, done according to the diner’s order at Haidilao's new artificial intelligence hotpot restaurant in Beijing, China, on November 14, 2018. — Reuters

  • Algorithm performs well in multiple US cities.
  • It studies city's criminal data and arrest history.
  • Previous efforts have been criticised for training algorithm on biased data.


According to a study published in the journal Nature Human Behaviour by researchers at the University of Chicago, a newly- developed algorithm has the ability to predict crimes a week prior to their occurrence with 90% accuracy.

Algorithms predict the weather, music taste, and whatnot. This new study, however, opens doors to a new room of possibilities for law enforcement agencies.

The system performed well in multiple US cities including Chicago, Los Angeles, Philadelphia, and San Fransisco, TRT World reported.

The algorithm studies a city's criminal data and arrest history, recognising the patterns in time and locations.

The study found that crimes in richer towns caused more arrests. This suggested discrimination in the police's attitude towards poorer neighbourhoods.

Previous similar efforts have been criticised for training the algorithm on biased data. 

This time, however, researcher are hoping that their findings will help law enforcement officials "hold states accountable in ways inconceivable in the past" as their study analyses biases.