Abstract:We consider analyzing traffic accident patterns using both road network data and satellite images aligned to road graph nodes. Previous work for predicting accident occurrences relies primarily on road network structural features while overlooking physical and environmental information from the road surface and its surroundings. In this work, we construct a large multimodal dataset spanning six U.S. states, containing nine million traffic accident records from official sources, and one million high-resolution satellite images for each node of the road network. Additionally, every node is annotated with features such as the region's weather statistics and road type (e.g., residential vs. motorway), and each edge is annotated with traffic volume information (i.e., Average Annual Daily Traffic). Utilizing this dataset, we conduct a comprehensive evaluation of multimodal learning methods that integrate both visual and network embeddings. Our findings show that integrating both data modalities improves prediction accuracy, achieving an average AUROC of $90.1\%$, a $3.7\%$ gain over graph neural network models that use only graph structures. With the improved embeddings, we conduct a causal analysis using a matching estimator to identify the key factors influencing traffic accidents. We find that accident rates rise by $24\%$ under higher precipitation, by $22\%$ on higher-speed roads such as motorways, and by $29\%$ due to seasonal patterns, after adjusting for other confounding factors. Ablation studies confirm that satellite imagery features are essential for achieving accurate prediction.
| Comments: | 17 pages. Appeared in KDD 2026 |
| Subjects: | Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Social and Information Networks (cs.SI) |
| Cite as: | arXiv:2512.02920 [cs.LG] |
| (or arXiv:2512.02920v3 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2512.02920 arXiv-issued DOI via DataCite |
|
| Related DOI: | https://doi.org/10.1145/3770854.3785677
DOI(s) linking to related resources |
Submission history
From: Ziniu Zhang [view email]
[v1]
Tue, 2 Dec 2025 16:39:32 UTC (39,670 KB)
[v2]
Wed, 17 Dec 2025 23:20:06 UTC (39,699 KB)
[v3]
Wed, 13 May 2026 18:31:48 UTC (39,699 KB)