Abstract:Traffic flow prediction plays a crucial role in alleviating traffic congestion and enhancing transport efficiency. While combining graph convolution networks with recurrent neural networks for spatial-temporal modeling is a common strategy in this realm, the restricted structure of recurrent neural networks limits their ability to capture global information. For spatial modeling, many prior studies learn a graph structure that is assumed to be fixed and uniform at all time steps, which may not be true. This paper introduces a novel traffic prediction framework, Global-Aware Enhanced Spatial-Temporal Graph Recurrent Network (GA-STGRN), comprising two core components: a spatial-temporal graph recurrent neural network and a global awareness layer. Within this framework, three innovative prediction models are formulated. A sequence-aware graph neural network is proposed and integrated into the Gated Recurrent Unit (GRU) to learn non-fixed graphs at different time steps and capture local temporal relationships. To enhance the model's global perception, three distinct global spatial-temporal transformer-like architectures (GST^2) are devised for the global awareness layer. We conduct extensive experiments on four real traffic datasets and the results demonstrate the superiority of our framework and the three concrete models.

What problem does this paper attempt to address?

This paper attempts to solve several key problems in traffic flow prediction: 1. **Limitations of Static Graph Structures**: Most existing traffic prediction models use fixed graph structures to model spatial relationships, and this assumption may not hold in practical applications. In fact, the graph structure of the road network changes over time, so a method that can adapt to dynamic graph structures is required. 2. **Capturing Spatio - Temporal Dependencies**: Although existing spatio - temporal graph recurrent neural networks (STGRNs) perform well in local spatio - temporal modeling, they are insufficient in capturing long - range temporal dependencies and highly dynamic spatial features. This limits their prediction performance in complex traffic scenarios. 3. **Global Information Perception Ability**: Existing STGRNs mainly focus on enhancing the spatial modeling part, and the internal recurrent operations make it difficult for them to capture the global temporal and spatial correlations in traffic data. To overcome these problems, the paper proposes a new traffic prediction framework - the Global - Awareness - Enhanced Spatio - Temporal Graph Recurrent Network (GA - STGRN). This framework contains two core components: - **Spatio - Temporal Graph Recurrent Neural Network**: It is used to learn non - fixed graph structures at different time steps and capture local temporal relationships. - **Global - Awareness Layer**: Three different global spatio - temporal transformer architectures (GST2) are designed to enhance the global perception ability of the model. Through these innovations, the GA - STGRN framework aims to improve the accuracy of traffic flow prediction, especially when dealing with complex and dynamic traffic data. The paper verifies the effectiveness of this framework through extensive experiments on four real - world traffic datasets, and the results show that GA - STGRN and its three specific models outperform existing baseline methods on multiple evaluation metrics.

Global-Aware Enhanced Spatial-Temporal Graph Recurrent Networks: A New Framework For Traffic Flow Prediction

A Spatial-Temporal Gated Hypergraph Convolution Network for Traffic Prediction

Attention-based Spatial-Temporal Graph Convolutional Recurrent Networks for Traffic Forecasting

STGMN: A gated multi-graph convolutional network framework for traffic flow prediction

Traffic flow prediction based on graph convolutional networks with a parallel attention network and stacked gate recurrent units

A Novel Attention-Based Dynamic Multi-Graph Spatial-Temporal Graph Neural Network Model for Traffic Prediction

Traffic Flow Prediction via Spatial Temporal Graph Neural Network

ST-AGRNN: A Spatio-Temporal Attention-Gated Recurrent Neural Network for Traffic State Forecasting

Bi-GRCN: A Spatio-Temporal Traffic Flow Prediction Model Based on Graph Neural Network

Spatial-Temporal Convolutional Graph Attention Networks for Citywide Traffic Flow Forecasting

Spatial-temporal graph neural network based on gated convolution and topological attention for traffic flow prediction

Attention-Based Gated Recurrent Graph Convolutional Network for Short-Term Traffic Flow Forecasting

Spatial-Temporal Graph Neural Network Framework with Multi-source Local and Global Information Fusion for Traffic Flow Forecasting

Spatio-Temporal Graph Neural Network for Traffic Prediction Based on Adaptive Neighborhood Selection

Learning Traffic As a Graph: A Gated Graph Wavelet Recurrent Neural Network for Network-Scale Traffic Prediction

Spatial-Temporal Graph Attention Gated Recurrent Transformer Network for Traffic Flow Forecasting

Multi-time Scale Spatial-Temporal Graph Neural Network for Traffic Flow Prediction

Local-Global Spatial-Temporal Graph Convolutional Network for Traffic Flow Forecasting

Traffic Forecasting with Graph Spatial-Temporal Position Recurrent Network.

STGC-GNNs: A GNN-based Traffic Prediction Framework with a Spatial-Temporal Granger Causality Graph

Dynamic Spatial–temporal Graph Convolutional Recurrent Networks for Traffic Flow Forecasting