An interpretable predictive model for crime forecasting using Graph Neural Network

• Grant number: 22/03941-0
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: December 01, 2022
• End date: January 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Luis Gustavo Nonato
• Grantee: Priscylla Maria da Silva Sousa
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated research grant: 13/07375-0 - CeMEAI - Center for Mathematical Sciences Applied to Industry, AP.CEPID
• Associated scholarship(s): 23/05783-5 - Investigating the disagreement problem in local explanation methods, BE.EP.DR

Abstract

Urban crime is a huge problem that affects the safety and security of society, especially in big cities. So, it is crucial the creation of mechanisms that help public agents to prevent crimes occurrences and understand which factors influence crime activities. In this context, machine learning methods have been applied to crime forecasting. However, the existing studies do not explore the city's structure, like the street network, and use limited data. Graph Neural Network (GNN) is a deep learning model with a solid potential for crime forecasting. Nevertheless, just a few studies use GNNs for crime. Moreover, it is hard to explain GNN predictions. In this project, we will investigate and propose a GNN model that captures the dynamics of crime using the city's street network and relying on a set of external factors, such as meteorological and socioeconomic data, to enable a more extensive comprehensive. In addition, the model will be integrated into an explainability technique to make the prediction easy to understand.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

SILVA, PRISCYLLA; SILVA, CLAUDIO; NONATO, LUIS GUSTAVO. Exploring the Relationship Between Feature Attribution Methods and Model Performance. NEURIPS WORKSHOPS, 2020, v. 257, p. 10-pg., 2024-01-01. (23/05783-5, 22/09091-8, 22/03941-0)

SILVA, PRISCYLLA; GUARDIEIRO, VITORIA; BARR, BRIAN; SILVA, CLAUDIO; NONATO, LUIS GUSTAVO. Visagreement: Visualizing and Exploring Explanations (Dis)Agreement. IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, v. 31, n. 10, p. 14-pg., 2025-10-01. (23/05783-5, 22/03941-0, 22/09091-8)