Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MDAS-GNN: Multi-Dimensional Spatiotemporal GNN with Spatial Diffusion for Urban Traffic Risk Forecasting

Ziyuan Gao

TL;DR

MDAS-GNN tackles urban road safety forecasting by modeling three distinct risk dimensions—traffic safety, infrastructure, and environmental risk—using feature-specific spatial diffusion and a weekly-focused multi-head temporal attention scheme. The approach builds a multi-dimensional feature tensor and employs differentiated diffusion over a road-network graph, enabling dimension-aware spatiotemporal learning. Empirical results across three UK regions show superior predictive accuracy and temporal stability over strong baselines, with ablations confirming gains from integrating all risk dimensions and from differentiated diffusion strategies. The framework yields actionable risk maps and supports targeted safety interventions and infrastructure planning, offering a robust tool for data-driven transportation safety management.

Abstract

Traffic accidents represent a critical public health challenge, claiming over 1.35 million lives annually worldwide. Traditional accident prediction models treat road segments independently, failing to capture complex spatial relationships and temporal dependencies in urban transportation networks. This study develops MDAS-GNN, a Multi-Dimensional Attention-based Spatial-diffusion Graph Neural Network integrating three core risk dimensions: traffic safety, infrastructure, and environmental risk. The framework employs feature-specific spatial diffusion mechanisms and multi-head temporal attention to capture dependencies across different time horizons. Evaluated on UK Department for Transport accident data across Central London, South Manchester, and SE Birmingham, MDASGNN achieves superior performance compared to established baseline methods. The model maintains consistently low prediction errors across short, medium, and long-term periods, with particular strength in long-term forecasting. Ablation studies confirm that integrated multi-dimensional features outperform singlefeature approaches, reducing prediction errors by up to 40%. This framework provides civil engineers and urban planners with advanced predictive capabilities for transportation infrastructure design, enabling data-driven decisions for road network optimization, infrastructure resource improvements, and strategic safety interventions in urban development projects.

MDAS-GNN: Multi-Dimensional Spatiotemporal GNN with Spatial Diffusion for Urban Traffic Risk Forecasting

TL;DR

MDAS-GNN tackles urban road safety forecasting by modeling three distinct risk dimensions—traffic safety, infrastructure, and environmental risk—using feature-specific spatial diffusion and a weekly-focused multi-head temporal attention scheme. The approach builds a multi-dimensional feature tensor and employs differentiated diffusion over a road-network graph, enabling dimension-aware spatiotemporal learning. Empirical results across three UK regions show superior predictive accuracy and temporal stability over strong baselines, with ablations confirming gains from integrating all risk dimensions and from differentiated diffusion strategies. The framework yields actionable risk maps and supports targeted safety interventions and infrastructure planning, offering a robust tool for data-driven transportation safety management.

Abstract

Traffic accidents represent a critical public health challenge, claiming over 1.35 million lives annually worldwide. Traditional accident prediction models treat road segments independently, failing to capture complex spatial relationships and temporal dependencies in urban transportation networks. This study develops MDAS-GNN, a Multi-Dimensional Attention-based Spatial-diffusion Graph Neural Network integrating three core risk dimensions: traffic safety, infrastructure, and environmental risk. The framework employs feature-specific spatial diffusion mechanisms and multi-head temporal attention to capture dependencies across different time horizons. Evaluated on UK Department for Transport accident data across Central London, South Manchester, and SE Birmingham, MDASGNN achieves superior performance compared to established baseline methods. The model maintains consistently low prediction errors across short, medium, and long-term periods, with particular strength in long-term forecasting. Ablation studies confirm that integrated multi-dimensional features outperform singlefeature approaches, reducing prediction errors by up to 40%. This framework provides civil engineers and urban planners with advanced predictive capabilities for transportation infrastructure design, enabling data-driven decisions for road network optimization, infrastructure resource improvements, and strategic safety interventions in urban development projects.

Paper Structure

This paper contains 30 sections, 25 equations, 9 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Multi-Dimensional Risk Assessment in Traffic Safety
  4. Spatiotemporal Modeling for Accident Prediction
  5. Spatial Dependencies and Propagation Mechanisms
  6. Temporal Dynamics and Multi-Scale Patterns
  7. Spatial Diffusion in Accident Analysis
  8. Three-Dimensional Risk Framework
  9. Framework Design
  10. Empirical Validation
  11. Methodology
  12. Data
  13. Temporal Aggregation Strategy
  14. Spatiotemporal Feature Learning Framework
  15. Multi-dimensional Feature Tensor Construction
  16. ...and 15 more sections

Figures (9)

  • Figure 1: Spatiotemporal network representation and temporal aggregation analysis, Sub-figure (a) illustrates road segment graph topology, Sub-figure (b) illustrates Spatial connectivity structure, Sub-figure (c) highlights Signal-to-noise ratio comparison across daily, weekly, and monthly aggregation granularities
  • Figure 2: Temporal risk evolution for three features
  • Figure 3: Spatial distribution for three features after iterated diffusion
  • Figure 4: MDAS-GNN: Model architecture
  • Figure 5: Temporal stability comparison
  • ...and 4 more figures