Unifying Geospatial Representations and Latent Crime Mapping from a Computational and Socio-Analytical Perspective

• Grant number: 23/15618-1
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: January 01, 2024
• End date: June 09, 2024
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Luis Gustavo Nonato
• Grantee: Thiago Rodrigo Ramos
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated research grant: 22/09091-8 - Criminality, insecurity, and legitimacy: a transdisciplinary approach, AP.ESCIENCE.TEM

Abstract

A recurring question in the analysis of the temporal evolution of crimes is the correlation between Urban Segments (SU), such as the industrial sector, central area, high, medium and low standard regions, and the most common Crime Types (TC), for example , homicides, thefts, among others. Another relevant issue in this context is to understand what the criminal evolution associated with each segment has been. The answers to such questions are too complex, since they are influenced not only by socioeconomic and cultural variables, but also by exogenous factors that are difficult to understand and systematize, such as, for example, the existence of the perception of differentiated punishment and the values that each individual attributed to relationship history and family formation. This postdoctoral project aims to combine SU mapping models in Remote Sensing images, spatial metrics, and deep graph neural networks (GNN - Graph Neutral Networks) to produce insecurity maps, thus addressing the questions raised above. The idea is to combine both representations: geospatial, obtained by satellite images, and latent, provided by the application of a model based on GNN, comparing each of this information in order to match segments and crimes. This approach will make it possible to innovate in the way of interpreting criminal dynamics by category, thus advancing the debate on how the social class of a given segment can influence the different types of crimes committed.