STG4Traffic: A Survey and Benchmark of Spatial-Temporal Graph Neural Networks for Traffic Prediction

TL;DR

Traffic prediction hinges on capturing spatial-temporal dependencies over road networks. This paper surveys graph-learning strategies and spatial-temporal graph neural networks (STGNNs), and introduces STG4Traffic, a PyTorch-based benchmark that standardizes data interfaces and evaluation across about 18 models on datasets such as METR-LA, PEMS-BAY, PEMSD4, and PEMSD8. It highlights that adaptive graph learning, diffusion-based convolutions, and attention mechanisms drive strong long-horizon forecasts, with methods like AGCRN offering efficiency and DAAGCN excelling on certain benchmarks; overall, STGNNs have improved prediction accuracy by roughly 15–20% over the past five years. The benchmark enables fair, reproducible comparisons and practical deployment considerations, while outlining key challenges—data quality, dynamic graphs, long-range dependencies, and cross-city transfer—that shape future research directions. The work thus provides a comprehensive foundation for advancing traffic forecasting through standardized evaluation and rigorous analysis of graph design and computation.

Abstract

Traffic prediction has been an active research topic in the domain of spatial-temporal data mining. Accurate real-time traffic prediction is essential to improve the safety, stability, and versatility of smart city systems, i.e., traffic control and optimal routing. The complex and highly dynamic spatial-temporal dependencies make effective predictions still face many challenges. Recent studies have shown that spatial-temporal graph neural networks exhibit great potential applied to traffic prediction, which combines sequential models with graph convolutional networks to jointly model temporal and spatial correlations. However, a survey study of graph learning, spatial-temporal graph models for traffic, as well as a fair comparison of baseline models are pending and unavoidable issues. In this paper, we first provide a systematic review of graph learning strategies and commonly used graph convolution algorithms. Then we conduct a comprehensive analysis of the strengths and weaknesses of recently proposed spatial-temporal graph network models. Furthermore, we build a study called STG4Traffic using the deep learning framework PyTorch to establish a standardized and scalable benchmark on two types of traffic datasets. We can evaluate their performance by personalizing the model settings with uniform metrics. Finally, we point out some problems in the current study and discuss future directions. Source codes are available at https://github.com/trainingl/STG4Traffic.

Figures (7)

• Figure 1: An example of different spatial-temporal correlations among routes. As shown in the left figure, spatial correlation is not only related to the distance, but also to the POI properties of the nodes, and the region similarity. The right figure presents the traffic trend of a node in two days, and the traffic pattern is basically the same for two consecutive days. Meanwhile, within one day, the traffic flows in the morning, afternoon and evening show significant variability.
• Figure 2: The Process of Discrete Graph Structure Learning.
• Figure 3: Architectures of Representative STGNN Models.
• Figure 4: Visualization of the MAE for each prediction step.
• Figure 5: Traffic speed or flow prediction at timeslots in one day on different datasets.