Scholarlink

<p>Urban traffic flow prediction is a fundamental task in intelligent transportation systems, yet it remains exceptionally challenging due to the entangled nature of spatial heterogeneity, non-stationary temporal dynamics, and multi-scale periodicity in real-world road networks. Existing graph neural network (GNN)- and transformer-based methods often treat spatial and temporal modeling as largely independent components, thereby overlooking the synergistic interactions between structural topology and sequential patterns. To address these limitations, we propose <b>ST-GNNFormer</b>, a novel <b>H</b>ybrid <b>S</b>patio-<b>T</b>emporal <b>G</b>raph <b>T</b>rans<b>f</b>ormer that tightly couples adaptive graph learning with multi-scale temporal attention for fine-grained traffic prediction. Specifically, ST-GNNFormer consists of four collaborating modules: (<i>i</i>)&#xa0;an <i>Adaptive Dynamic Graph Learning</i> (ADGL) module that infers time-varying adjacency from learnable node embeddings conditioned on temporal context, capturing both structural and semantic proximity; (<i>ii</i>)&#xa0;a <i>Spatial Graph Transformer</i> (SGT) that integrates graph-structure biases into multi-head self-attention to propagate spatially correlated features over the learned topology; (<i>iii</i>)&#xa0;a <i>Temporal Multi-Scale Transformer</i> (TMST) that simultaneously models short-term fluctuations and long-range periodicity across multiple temporal granularities; and (<i>iv</i>)&#xa0;a <i>Cross-Scale Fusion</i> (CSF) gate that adaptively merges spatial and temporal representations at each encoder layer. Extensive experiments on four public benchmarks—METR-LA, PEMS-BAY, PEMS04, and PEMS08—demonstrate that ST-GNNFormer consistently outperforms 8 competitive baselines by up to 7.5% in MAE at the 60-minute prediction horizon while maintaining competitive computational efficiency. Ablation studies and visualization analyses confirm the individual contributions of each module and reveal physically meaningful attention patterns that align with real-world traffic dynamics.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ST-GNNFormer: coupling dynamic graph learning and multi-scale temporal attention for traffic flow forecasting

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月1日

查看全文
  • Zhengjia Chen,
  • Junhao Chen

摘要

Urban traffic flow prediction is a fundamental task in intelligent transportation systems, yet it remains exceptionally challenging due to the entangled nature of spatial heterogeneity, non-stationary temporal dynamics, and multi-scale periodicity in real-world road networks. Existing graph neural network (GNN)- and transformer-based methods often treat spatial and temporal modeling as largely independent components, thereby overlooking the synergistic interactions between structural topology and sequential patterns. To address these limitations, we propose ST-GNNFormer, a novel Hybrid Spatio-Temporal Graph Transformer that tightly couples adaptive graph learning with multi-scale temporal attention for fine-grained traffic prediction. Specifically, ST-GNNFormer consists of four collaborating modules: (i) an Adaptive Dynamic Graph Learning (ADGL) module that infers time-varying adjacency from learnable node embeddings conditioned on temporal context, capturing both structural and semantic proximity; (ii) a Spatial Graph Transformer (SGT) that integrates graph-structure biases into multi-head self-attention to propagate spatially correlated features over the learned topology; (iii) a Temporal Multi-Scale Transformer (TMST) that simultaneously models short-term fluctuations and long-range periodicity across multiple temporal granularities; and (iv) a Cross-Scale Fusion (CSF) gate that adaptively merges spatial and temporal representations at each encoder layer. Extensive experiments on four public benchmarks—METR-LA, PEMS-BAY, PEMS04, and PEMS08—demonstrate that ST-GNNFormer consistently outperforms 8 competitive baselines by up to 7.5% in MAE at the 60-minute prediction horizon while maintaining competitive computational efficiency. Ablation studies and visualization analyses confirm the individual contributions of each module and reveal physically meaningful attention patterns that align with real-world traffic dynamics.