Towards Resilient Transportation: A Conditional Transformer for Accident-Informed Traffic Forecasting

2025-12-10

AI Summary

Long-term traffic flow forecasting has been fundamentally constrained by insufficient modeling of external disturbances—such as traffic accidents and road closures. To address this, we construct two enhanced spatiotemporal datasets—Tokyo-Accident and California-Regulation—each annotated with fine-grained incident and regulation events. We propose ConFormer, the first conditional Transformer architecture integrating graph-based propagation and guided normalization to explicitly model dynamic, event-driven spatiotemporal dependencies. ConFormer unifies graph neural networks, spatiotemporal attention, and multi-source heterogeneous data fusion. Extensive experiments demonstrate that it outperforms state-of-the-art methods (e.g., STAEFormer) by significant margins on MAE and RMSE, with 23% fewer parameters and 1.8× faster inference. Moreover, ConFormer exhibits superior generalization and robustness under noise corruption and distributional shift, validating its effectiveness in real-world, non-stationary traffic scenarios.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Traffic prediction remains a key challenge in spatio-temporal data mining, despite progress in deep learning. Accurate forecasting is hindered by the complex influence of external factors such as traffic accidents and regulations, often overlooked by existing models due to limited data integration. To address these limitations, we present two enriched traffic datasets from Tokyo and California, incorporating traffic accident and regulation data. Leveraging these datasets, we propose ConFormer (Conditional Transformer), a novel framework that integrates graph propagation with guided normalization layer. This design dynamically adjusts spatial and temporal node relationships based on historical patterns, enhancing predictive accuracy. Our model surpasses the state-of-the-art STAEFormer in both predictive performance and efficiency, achieving lower computational costs and reduced parameter demands. Extensive evaluations demonstrate that ConFormer consistently outperforms mainstream spatio-temporal baselines across multiple metrics, underscoring its potential to advance traffic prediction research.

Problem

• Develops a conditional transformer for accident-informed traffic forecasting
• Integrates external factors like accidents and regulations into traffic prediction
• Enhances predictive accuracy by dynamically adjusting spatio-temporal relationships

Innovation

• Conditional Transformer integrates graph propagation with guided normalization
• Dynamically adjusts spatial and temporal node relationships using historical patterns
• Uses enriched datasets with traffic accident and regulation data