Spatial Concentration Patterns of Vehicle Theft in Municipalities of Great João Pessoa Using Machine Learning Techniques

Abstract

Purpose: The objective of this paper is to identify patterns of spatial concentration related to robbery and theft of vehicles in the great João Pessoa (PB), a region composed by the municipalities of Bayeux, Cabedelo, João Pessoa and Santa Rita, using machine learning techniques. Thus, it seeks to contribute to the discussion on the potential benefits of using Artificial Intelligence tools in the field of public security. Methodology: The data used were obtained from the Secretaria de Estado da Segurança e da Defesa Social da Paraíba and include the years from 2017 to 2019. The base is made up of 5.385 occurrences of robbery and theft of cars and motorcycles, indicating municipality, neighborhood, day of week, shift and time of crime. The empirical strategy adopted consisted in the application of the Density-based Spatial Clustering of Applications with Noise (DBSCAN) algorithm. Findings: The descriptive analysis showed that the center of João Pessoa is the neighborhood with the highest rate of vehicle subtraction per 100 thousand inhabitants, followed by Barra de Gramame, Ponta do Seixas, Distrito Industrial and Varadouro. In relation to the concentration of crimes in certain places, the use of the DBSCAN algorithm allowed to identify hotspots for different days and shifts, being that the number of these was higher during the weekdays at night. Contributions: These results have the potential to help develop a more effective public security planning in the neighborhoods of the great João Pessoa, as they suggest how to move the police force in order to achieve greater efficiency in preventing crimes and capturing criminals.