A new algorithm can predict the occurrence of crimes in large cities, using temporal and geographic patterns, according to researchers at the University of Chicago, in the United States. They claim that predictions are about 90% accurate and can be made up to a week before the fact.
The team responsible for creating the algorithm is made up of data scientists and social scientists. The initiative was revealed in a study published in the scientific journal Nature.
Despite sounding like science fiction, the researchers say it’s not the kind of solution used in the movie “Minority Report” (2002), in which police officers were able to visualize exactly the type of crime and the perpetrator. The main point of the system is to make predictions based on criminal data.
The code was fed public data on violent crimes (such as homicides, beatings and assaults) and property crimes (such as robberies and thefts). From this, they were able to identify patterns to make predictions.
Crime knows no bounds
Police forces had already tried to use other models to attempt to predict the occurrence of crimes. But, according to the researchers, their algorithm is more efficient because the old systems identified areas where crimes were recorded and predicted that they would spread to neighboring regions from there.
The new divides the cities into equal areas of about 93 square meters. This is important because other divisions, such as neighborhood boundaries, are ignored, seeking to eliminate the political and social bias that exists in this type of demarcation.
“The spatial models ignore the natural topology of the city,” said sociologist James Evans, one of the co-authors, in an interview with the university’s official website. “Transport networks respect streets, sidewalks and train and bus lines. Communication networks respect areas of similar socioeconomic background. Our model allows for the discovery of these connections.”
The study shows that in addition to Chicago (Illinois), accurate predictions were made for other major US cities such as Atlanta (Georgia), Los Angeles (California), Portland (Oregon) and San Francisco (California). In all of them, the performance of the algorithm was satisfactory.
Rich neighborhood or poor neighborhood?
Another interesting aspect is that the study takes into account that the police response to crime varies according to neighborhoods. Basically, crimes in places with higher socioeconomic status result in more arrests. In less developed neighborhoods, the number of arrests remains unchanged, even with the increase in crime.
According to another researcher on the study, Ishanu Chattopadhyay, the phenomenon demonstrates how crime in the richest areas makes the answer to increase security in them, which takes away policing resources from the poorest regions.
However, he cautions: the algorithm should not be used to simply direct police action, as in the “Minority Report”. His recommendation is that this can be an additional tool in urban policing strategies.
“You can use it as a simulation tool to see what happens when crime occurs in one area of the city, or when there is increased security enforcement in another. If you apply all these variables, you can see how the system responds.”