Data Integration and Crime Hotspot Prediction

• Grant number: 24/15983-4
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: November 01, 2024
• End date: October 31, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Luis Gustavo Nonato
• Grantee: Samuel de França Marques
• 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

Existing techniques for the crime analysis allow only rough mechanisms to investigate the different facets of illegal activities. For example, extracting, grouping and visualizing crime pattern over time is a hard task to accomplish using the existing techniques. In addition, the existing solutions are adapted to reveal crime hotspots based on the number of crimes and not necessarily on the crime frequency, thus limiting the analysis scope and prediction capability. We propose the development of new tools which are able to identify and predict hotspots based not only on the crime intensity, but also on the probability of crime occurrences. Combining the intensity of hotspots with their probability of occurrence allows new alternatives of analysis, making it possible to reveal and predict crime patterns depending on the location and time period. In addition, we propose the use of deep learning, more specifically Graph Neural Networks (GNN), as the main mechanism to predict crime activities. Deep learning updated tools will be used to map crime time series in latent spaces. Such tools will support visualization tools which are able to deal with spatial and temporal patterns simultaneously.