The records of complaints, data on crimes committed and other data on the incidence of crime They make up a source of information with possibilities of being valued. As? Well, at the service of crime preventionwith a mathematical model that predicts the incidence of crime and that it will help develop more effective anti-crime policies.
This has been demonstrated by the National University of Colombia (UNAL), which has applied machine learning techniques in an artificial intelligence toolwhich can be used to identify patterns of crime occurrence in space and time.
These patterns allow predictions about future occurrences of crimes, determine common behaviors in the phenomena, and even quantify subjective perceptionssuch as fear of crime, deeply related to the perception of security.
Professor Francisco Albeiro Gómez Jaramillo, from the Department of Mathematics at UNAL, principal investigator of the project “Design and validation of predictive analytical models of security and coexistence phenomena for decision-making in Bogotá”, explains that “lPrediction models seek to anticipate and explain the occurrence of security phenomena using historical data from the reports of these crimes.”
“For example, this approximation helps to understand where and when these phenomena could most likely occur and give clues about the possible reasons for their occurrence.”
Security perception indicator
The data used to develop the tool correspond to more than 3 million complaints made by citizens (for robbery, fights and homicides, among others), calls to line 123 and subpoenas imposed by the Police.
These records feed artificial intelligence models that characterize the patterns of occurrence. Thus, it was possible to identify that “in the towns of Ciudad Bolívar, Usme, Bosa and Suba, fights are more predictable and tend to happen especially on weekends.”
On the other hand, in the town of Teusaquillo there are fights almost every day, but the pattern is less predictable. Possibly these schemes occur due to the daily routines of the city, such as nightlife in the downtown areas.
Another result of the project was the construction of a security perception indicatordeveloped from more than 1,700,000 tweets(status messages on Twitter that can have up to 280 characters) published in Bogotá and collected for more than a year (March 18, 2019 – April 28, 2020).
“In this case, the artificial intelligence models automatically identified the messages related to security, and then quantified between 1 and 5 the level of positive perception of security through patterns extracted from annotations made by experts on the subject,” explains the teacher.
Predict the number of crime-related social media posts
With these results, “together with information on temporary events -such as protests, soccer matches and special dates such as Mother’s Day or the New Year- the models predicted the number of tweets related to security that could appear in the near future, and resulted in a more general indicator of the perception of security.
Due to the importance of its contributions, the project was nominated among the top three in the “Innovation” category of the 2022 Well-Invested Royalties Awards, for being an innovative response to the social, economic or environmental problems or needs of the city and the country.
Among the results of the project are the publication of four articles in specialized magazines and the presentation of innovations in more than 20 conferences at international scientific events.
In addition, one of the scientific contributions was awarded as the best student article at the Seventh International Conference on Behavioral and Social Computing (BESC 2020), held in Bournemouth (United Kingdom), and another contribution was recognized with the International Charles François Award from the Academy International of Systems and Cybernetic Sciences.