Graph neural networks (GNNs) are a powerful tool for handling spatial dependence. In this paper we present how they can be used to estimate the propensity to get an accident for any edge of a road network. The results might be useful for policy makers to mitigate crash events. Implementing GNN requires some care due to the strong imbalance of the dataset. A test to validate the capability of GNN to capture the local dependence of crashes has been considered.
Cantaluppi, G., Lin, J., Valena, G., Zappa, D., Graph Neural Networks for Traffic Accident Risk, in Boccuzzo, G., Bovo, E., Manisera, M., Salmaso, L. (ed.), IES 2025 - Innovation & Society: Statistics and Data Science for Evaluation and Quality. BOOK OF SHORT PAPERS, Cleup, Padova 2025: 959- 965 [https://hdl.handle.net/10807/318036]
Graph Neural Networks for Traffic Accident Risk
Cantaluppi, Gabriele;Lin, Jianyi;Valena, Gianni;Zappa, Diego
2025
Abstract
Graph neural networks (GNNs) are a powerful tool for handling spatial dependence. In this paper we present how they can be used to estimate the propensity to get an accident for any edge of a road network. The results might be useful for policy makers to mitigate crash events. Implementing GNN requires some care due to the strong imbalance of the dataset. A test to validate the capability of GNN to capture the local dependence of crashes has been considered.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
