ST-RetNet: A Long-term Spatial-Temporal Traffic Flow Prediction Method

TL;DR

ST-RetNet tackles the challenge of long-term traffic flow forecasting on spatial-temporal data by introducing a dual-branch RetNet-based framework. It fuses S-RetNet, which combines static spatial cues from a graph convolutional network with a learnable adaptive adjacency, and T-RetNet, which leverages temporal retention for long-range dependencies, enabling parallel processing. The model expands input features, applies multi-scale spatial-temporal retention blocks, and uses a two-stage prediction head, achieving superior accuracy on four real-world datasets compared to state-of-the-art baselines. The approach offers practical impact for intelligent transportation systems by delivering accurate, scalable, and parsimonious long-horizon traffic forecasts, with potential for multimodal extensions and integration with language-model-inspired components in the future.

Abstract

Traffic flow forecasting is considered a critical task in the field of intelligent transportation systems. In this paper, to address the issue of low accuracy in long-term forecasting of spatial-temporal big data on traffic flow, we propose an innovative model called Spatial-Temporal Retentive Network (ST-RetNet). We extend the Retentive Network to address the task of traffic flow forecasting. At the spatial scale, we integrate a topological graph structure into Spatial Retentive Network(S-RetNet), utilizing an adaptive adjacency matrix to extract dynamic spatial features of the road network. We also employ Graph Convolutional Networks to extract static spatial features of the road network. These two components are then fused to capture dynamic and static spatial correlations. At the temporal scale, we propose the Temporal Retentive Network(T-RetNet), which has been demonstrated to excel in capturing long-term dependencies in traffic flow patterns compared to other time series models, including Recurrent Neural Networks based and transformer models. We achieve the spatial-temporal traffic flow forecasting task by integrating S-RetNet and T-RetNet to form ST-RetNet. Through experimental comparisons conducted on four real-world datasets, we demonstrate that ST-RetNet outperforms the state-of-the-art approaches in traffic flow forecasting.

Figures (4)

• Figure 1: The parallel representation of traffic flow forecasting.
• Figure 2: The framework of ST-RetNet.
• Figure 3: Performance comparison between S-RetNet and GCN.
• Figure 4: Performance comparison between T-RetNet and temporal models.