A new computer-based algorithm by social scientists at the University of Chicago (USA) can now predict crime in a large city near where people live, with 90% accuracy.
$ads={1}
According to a study published in the journal Nature Human Behavior, the algorithm divided the city of Chicago into 93-square-foot squares, then researchers used historical data on violent crime and crime. City assets to check in each block.Through the information collected, the new algorithm is able to predict possible crimes a week in advance in the area, Bloomberg News reported.
The study found that when using data from other major cities, including Atlanta, Los Angeles and Philadelphia, the new algorithm also gave fairly accurate predictions.
The new tool contrasts with previous predictive models, which mainly depict crime emerging from "hot spots" and spreading to surrounding areas.
The old model's approach, according to the report, tends to overlook the complex social environment of cities, as well as the nuanced relationship between crime and the impact of police enforcement. at the same time easy to confuse.
Crime prediction models previously used by many law enforcement agencies have been found to mistarget certain audiences because they are based on a narrower set of factors.
In 2012, the Chicago Police Department and academic researchers developed the "Risk Crime and Victim Model" to generate a list of the people of interest, or victims and perpetrators of crimes. potential shooter. All are determined from factors such as age and arrest history. The release of this list caused many lawsuits.
After a lengthy legal battle, however, an investigation by the Chicago Sun-Times in 2017 revealed that nearly half of the people the model identified as potential perpetrators had never been charged with a crime. illegal gun possession, of which 13% have never been charged with a felony.
In contrast, the new tool, designed by study author Ishanu Chattopadhyay and colleagues at the University of Chicago, uses hundreds of thousands of sociological data to find out the risk of committing a crime at a time and without specific time, reaching the required accuracy up to 90%.
The study, titled "Predicting Urban Crime and Revealing Crime Trends in American Cities," was conducted by the Defense Advanced Research Projects Agency and the College of Culture and Society. Supported by the Neubauer Society.