Menu Close

A new algorithm developed in the US can predict crimes a week ahead

A new algorithm developed in the US can predict crimes a week ahead

A new algorithm developed in the US can predict crimes a week ahead. Scientists at the University of Chicago have developed an algorithm that can predict crime in urban areas a week ahead. This is according to a new study published in Nature Human Behavior. The algorithm was found to predict crimes in a range of a thousand feet with 90% accuracy.

The algorithm learned patterns from publicly available data on violent and property crimes in order to do this. This is according to the scientists. The authors say, “We reveal a way to predict crime in cities at the level of individual incidents with a level of accuracy that is much higher than what has been achieved in the past.”


The prevalence of algorithms has dramatically increased during the last few years. Algorithms are used in everything from weather forecasting to driving automobiles, providing shopping suggestions and discovering medical treatments. So it’s not surprising that it’s being used to combat crime.

The algorithm was tested and verified by utilizing historical data from the City of Chicago. This is about two major categories of recorded events—violent crimes (homicides, assaults, and battery) and property crimes (burglaries, thefts, and motor vehicle thefts).

Because people in cities haven’t always trusted or worked with the police, these statistics were used because they were more likely to be reported to the police in those places. In contrast to drug offenses, traffic stops and other minor violations, these offenses are also less likely to be subject to enforcement prejudice.

A new algorithm developed in the US can predict crimes a week ahead

This raised the question: how does the model stay impartial? Well, the scientists said the new model separates crimes by looking at when and where they happened. It also looks for patterns to predict when more crimes will happen.

What is the new algorithm’s procedure?

The tool splits a city into 1,000 square feet tiles and forecasts future occurrences using previous data on violent and property crimes. According to the researchers, this method is unique compared to previous algorithmic forecasts. This is because it views crime as developing from hotspots and spreading to other places.

However, social scientists contend that such methods overlook the diverse social environments of cities. It is also skewed by the state-run monitoring that is employed for law enforcement.

Instead, the algorithm looked at past crime records and took a number of other things into account. It then predicted the likelihood of crime in Chicago to within 90% accuracy.

In addition, the algorithm was tested to see how well it predicted crimes in eight different American cities. This includes Los Angeles, Atlanta, and Philadelphia, and it performed admirably in all of them with 90 percent accuracy.

Before the Olympics, the Tokyo Police Department wanted to see if they could use AI-based technologies to foresee crimes before they happened. Even while it may seem like we already inhabit a Minority Report-style future, this has really been the case for about a decade.

Crime and Victimization Risk Model

In 2012, the Chicago Police Department put the Crime and Victimization Risk Model into practice. This was done with the help of several academic experts.

The program created a list of possible attackers and their victims based on criteria like age and criminal records. It also gave each person on the list a score. This was to help law enforcement agencies determine how urgent it was to find the projected offender and their victim.

Although the idea could be intriguing, the execution was questionable. As further investigations revealed, roughly half of the accused offenders on the list had never been charged with unlawful possession of guns, and others had not previously been charged with major crimes.


A 2019 study from Technology Review found that the risk assessment algorithms that decide if someone should go to jail were trained on skewed data from the past.

In order to avoid making the same mistakes twice, University of Chicago scientists, led by assistant professor Ishanu Chattopadhyay set out to create their algorithm.

Previous models focused more on bias-prone conventional political or local borders. But, with data from seven other American cities, including Atlanta, Austin, Detroit, Los Angeles, Philadelphia, Portland, and San Francisco, the new model performed well.

How the algorithm should be used

Ishanu Chattopadhyay, the lead author, is quick to point out that despite the tool’s accuracy, it should not be used to direct law enforcement strategy. For instance, police agencies shouldn’t use it to pro-actively swarm communities to reduce crime, he said.

Instead, it needs to be included in a toolkit of law enforcement and urban policy measures to combat crime. “We built digital twins of urban settings. It will tell you what will happen in the future if you feed it data from what has already occurred. It is not magic,” he said.

We validated it and it works extremely well. Now you can use this as a simulation tool to examine what occurs if crime increases in one part of the city or enforcement is stepped up in another part of the city,” he added.

You may observe how the systems adapt when you apply all these various factors. But there is still a problem with possible policing prejudice.

Check out: Aging will soon be a choice as scientists close in on the fountain of youth

Law Enforcement

The study team also examined how quickly police make arrests after crimes occur. It compared those rates across o ther communities to learn more about the police response to crime.


They discovered that more arrests occurred when crime levels rose in affluent neighborhoods. But this didn’t happen in poor areas, which shows that police enforcement and response aren’t the same everywhere.

You may like: New technology makes heart repair and regeneration simple

In their conclusion, the authors say, “We acknowledge the danger that powerful predictive tools put in the hands of overzealous states in the name of civilian protection. However, here we show their unprecedented ability to audit enforcement biases and hold states accountable in ways that were unthinkable in the past.”

The study “Event-level Prediction of Urban Crime Reveals Signature of Enforcement Bias in U.S. Cities” was authored by Victor Rotaru, Yi Huang, Timmy Li, James Evans and Ishanu Chattopadhyay.

Leave a Reply