Urban Traffic Accident Risk Prediction Revisited: Regionality, Proximity, Similarity and Sparsity

TL;DR

The paper tackles urban traffic accident risk prediction by addressing regional background, spatial proximity, semantic similarity, and data sparsity. It introduces the multi-granularity hierarchical spatio-temporal network (MGHSTN), which integrates remote sensing data, a dual hierarchical data construction, and multi-view semantic graphs, coupled with adaptive temporal attention and a multivariate loss to jointly learn region and graph representations. Empirical results on NYC and Chicago datasets show substantial improvements over state-of-the-art baselines in RMSE, Recall, and MAP, with notable robustness to zero-inflation and enhanced interpretability through regional background cues. The approach offers practical impact for city planning and traffic safety by providing more accurate, robust, and region-aware risk predictions.

Abstract

Traffic accidents pose a significant risk to human health and property safety. Therefore, to prevent traffic accidents, predicting their risks has garnered growing interest. We argue that a desired prediction solution should demonstrate resilience to the complexity of traffic accidents. In particular, it should adequately consider the regional background, accurately capture both spatial proximity and semantic similarity, and effectively address the sparsity of traffic accidents. However, these factors are often overlooked or difficult to incorporate. In this paper, we propose a novel multi-granularity hierarchical spatio-temporal network. Initially, we innovate by incorporating remote sensing data, facilitating the creation of hierarchical multi-granularity structure and the comprehension of regional background. We construct multiple high-level risk prediction tasks to enhance model's ability to cope with sparsity. Subsequently, to capture both spatial proximity and semantic similarity, region feature and multi-view graph undergo encoding processes to distill effective representations. Additionally, we propose message passing and adaptive temporal attention module that bridges different granularities and dynamically captures time correlations inherent in traffic accident patterns. At last, a multivariate hierarchical loss function is devised considering the complexity of the prediction purpose. Extensive experiments on two real datasets verify the superiority of our model against the state-of-the-art methods.

Figures (9)

• Figure 1: Critical Aspects of Traffic Accident.
• Figure 2: An illustration of preliminaries, (a) shows the region features input and (b) shows the graph features input.
• Figure 3: The model framework of MGHSTN.
• Figure 4: The Framework of Remote Sensing Autoencoder
• Figure 5: Multi-level Embedding Fusion Mechanism