Uma nova metodologia baseada em grafos de ruas para investigar a relação entre crimes e pontos de interesse usando dados georreferenciados: um estudo de caso em Maceió, AL

Abstract

Just like cities, crimes have also evolved over time: they are exponentially more intense, more violent, and more modern, leading to the exhaustion of security models. Consequently, society, and especially managers, need sophisticated tools to assist them in decision-making. The increasing digitization of data in the last decade has enabled the large-scale and agile collection of urban data. This opens up opportunities to develop new techniques and tools for analyzing massive urban data, including within the scope of urban crime. However, the analysis of large-scale georeferenced urban data requires the use of appropriate spatial discretizations and the employment of robust Machine Learning algorithms capable of identifying complex urban patterns. In this work, we present a computational methodology based on street graphs to perform the analysis of massive georeferenced urban data with the aim of investigating the relationship between the occurrence of crimes and the proximity of points of interest (POIs) in the city. In particular, we also propose to carry out segmented analyses according to the socioeconomic patterns of different city regions using clustering algorithms. Through a case study conducted in the city of Maceió, we conclude that there is, globally, a correlation between points of interest and criminal events. Moreover, this correlation is significantly altered when analyzing groups of corners segmented according to different socioeconomic patterns.