Embedding spatial context in urban traffic forecasting with contrastive pre-training

2025-03-19

AI Summary

To address insufficient spatial contextual modeling and underutilization of non-traffic data (e.g., road geometry) in urban traffic forecasting, this paper proposes the Traffic Quotient Graph (TQG) paradigm, unifying road topology, geometric attributes, and sensor observations into a single structured representation. We design an OpenStreetMap-enhanced geometric encoder and introduce a geography-aware contrastive self-supervised pretraining strategy, significantly improving model generalization without requiring additional traffic data. The TQG seamlessly integrates into graph neural network–based spatiotemporal forecasting frameworks, yielding substantial accuracy gains—especially for cross-region transfer and unseen road segments with no historical observations. Key contributions include: (i) the first structured spatial-semantic modeling of traffic via the TQG; (ii) a novel geography-informed contrastive pretraining methodology; and (iii) a new forecasting paradigm achieving high generalizability with minimal reliance on task-specific traffic data.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Computer Vision: Remote Sensing / Geospatial AI

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs 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

Urban traffic forecasting is a commonly encountered problem, with wide-ranging applications in fields such as urban planning, civil engineering and transport. In this paper, we study the enhancement of traffic forecasting with pre-training, focusing on spatio-temporal graph methods. While various machine learning methods to solve traffic forecasting problems have been explored and extensively studied, there is a gap of a more contextual approach: studying how relevant non-traffic data can improve prediction performance on traffic forecasting problems. We call this data spatial context. We introduce a novel method of combining road and traffic information through the notion of a traffic quotient graph, a quotient graph formed from road geometry and traffic sensors. We also define a way to encode this relationship in the form of a geometric encoder, pre-trained using contrastive learning methods and enhanced with OpenStreetMap data. We introduce and discuss ways to integrate this geometric encoder with existing graph neural network (GNN)-based traffic forecasting models, using a contrastive pre-training paradigm. We demonstrate the potential for this hybrid model to improve generalisation and performance with zero additional traffic data. Code for this paper is available at https://github.com/mattchrlw/forecasting-on-new-roads.

Problem

• Enhancing urban traffic forecasting using spatio-temporal graph methods.
• Incorporating spatial context to improve traffic prediction performance.
• Developing a hybrid model with contrastive pre-training for better generalization.

Innovation

• Traffic quotient graph combines road and traffic data
• Geometric encoder pre-trained with contrastive learning
• Integration with GNN models enhances forecasting accuracy