Graph signal processing and deep learning for crime prediction in São Paulo City

Abstract

São Paulo is the largest city in South America, with criminality rates as large as the city. The number and type of crimes in São Paulo vary considerably depending on urban, environmental, and social characteristics of locality. Previous studies about crime analysis in S ao Paulo city have mostly focused on uncovering crime patterns associated with socioeconomic factors, time seasonality, and urban routine activities. Moreover, those studies and associated tools have been designed to analyze the city as a whole, not being appropriated to study particular regions such as specific neighborhoods, streets, avenues, and parks. There is still another important aspect in this context, the availability of relevant information as to mobility, passersby behavior, urban infrastructure, and social media, which are not taken into account by existing crime pattern analysis tools. In this work, we propose the development of an analytic tool to analyze and predict crime patterns in specific regions of a big city as S ao Paulo. Our focus is on the identification and prediction of local hotspots, as well as their corresponding crime patterns. The idea is to exploit information obtained from to urban infrastructure and social network data to feed a deep neural network, which should learn the crime patterns. As a result, we will build an application which can be used by the population, increasing transparency of criminality in specific neighborhoods while helping government security agencies in their planning and studies. (AU)