MAF-GNN: Multi-adaptive Spatiotemporal-flow Graph Neural Network for Traffic Speed Forecasting

Abstract

Traffic forecasting is a core element of intelligent traffic monitoring system. Approaches based on graph neural networks have been widely used in this task to effectively capture spatial and temporal dependencies of road networks. However, these approaches can not effectively define the complicated network topology. Besides, their cascade network structures have limitations in transmitting distinct features in the time and space dimensions. In this paper, we propose a Multi-adaptive Spatiotemporal-flow Graph Neural Network (MAF-GNN) for traffic speed forecasting. MAF-GNN introduces an effective Multi-adaptive Adjacency Matrices Mechanism to capture multiple latent spatial dependencies between traffic nodes. Additionally, we propose Spatiotemporal-flow Modules aiming to further enhance feature propagation in both time and space dimensions. MAF-GNN achieves better performance than other models on two real-world datasets of public traffic network, METR-LA and PeMS-Bay, demonstrating the effectiveness of the proposed approach.

Spatial-Temporal Feature Extraction and Evaluation Network for Citywide Traffic Condition Prediction

“…Notable research includes that of Zheng et al, who integrated both spatial and temporal attention mechanisms to concurrently extract spatiotemporal features, later merging them through a gating fusion module [11]. Xu et al devised spatial and time transformer modules grounded on the attention mechanism [15]. The spatial transformer utilized the selfattention mechanism to capture real-time traffic conditions and traffic flow directionality.…”

“…As GCN evolved, fusion schemes of traffic predictors based on GCN have emerged. For instance, combinations of GCN with CNN have been proposed to capture spatial-temporal traffic data characteristics [14][15][16], and GCN with RNN have been used to separately model spatial and temporal features [10,17]. As attention mechanisms advanced in extracting sequence element correlations, strategies utilizing attention to capture spatial and temporal correlations have developed [18,19].…”

Spatial Downscaling of Streamflow Data with Attention Based Spatio-Temporal Graph Convolutional Networks

“…However, when temporal correlations are present, a standard GNN approach does not offer significant advantages. Forecasts or estimates considering spatial and temporal dependencies have been a challenge not only in streamflow but in other domains as well, such as traffic forecasting (Li et al, 2023;Tao et al, 2023) or traffic speed forecasts (Xu et al, 2021;Zhang et al, 2021). Recently, many deeplearning approaches have been proposed to overcome similar challenges, and models that are based on graph convolutional neural networks with additional components to capture a better understanding of complex spatial and temporal relationships have gained lots of attention (Jiang and Luo, 2022;Wu et al, 2020).…”