Spatial‐temporal attention wavenet: A deep learning framework for traffic prediction considering spatial‐temporal dependencies

Abstract

Traffic prediction on road networks is highly challenging due to the complexity of traffic systems and is a crucial task in successful intelligent traffic system applications. Existing approaches mostly capture the static spatial dependency relying on the prior knowledge of the graph structure. However, the spatial dependency can be dynamic, and sometimes the physical structure may not reflect the genuine relationship between roads. To better capture the complex spatial‐temporal dependencies and forecast traffic conditions on road networks, a multi‐step prediction model named Spatial‐Temporal Attention Wavenet (STAWnet) is proposed. Temporal convolution is applied to handle long time sequences, and the dynamic spatial dependencies between different nodes can be captured using the self‐attention network. Different from existing models, STAWnet does not need prior knowledge of the graph by developing a self‐learned node embedding. These components are integrated into an end‐to‐end framework. The experimental results on three public traffic prediction datasets (METR‐LA, PEMS‐BAY, and PEMS07) demonstrate effectiveness. In particular, in the 1 h ahead prediction, STAWnet outperforms state‐of‐the‐art methods with no prior knowledge of the network.

SDGNet: A Handover-Aware Spatiotemporal Graph Neural Network for Mobile Traffic Forecasting

“…• STAWnet [17] -introduces self-learned node embedding by a dynamic attention mechanism; no prior knowledge of the graph is needed; the adjacency matrix is not used in attention learning. All models considered are trained and evaluated with the same dataset.…”

Multivariate and Propagation Graph Attention Network for Spatial-Temporal Prediction with Outdoor Cellular Traffic

“…Comparison Methods: Our dataset is a spatial-temporal dataset with a road network, similar tasks such as traffic speed have generally been addressed better by graph-based models [20]. Due to the page limitation, we focus on the latest graph-based models as baselines in our evaluation: LSTM [5], GWENT [17], MTGNN [16], STAWNET [10]. To verify the effectiveness of the correlation among multivariate, we build two models, MPGAT-1 and MPGAT, where MPGAT-1 only adopt univariate IMEI quantity as input.…”

“…Spatial-temporal prediction with a road network has been widely studied. Recent researches [8,10,16,17] have achieved great success in prediction by conducting propagating information on the graph data within graph neural networks (GNN). However, the dataset of previous studies is the traffic speed detected from sensors, which is usually less varied between time steps.…”

STDEN: Towards Physics-Guided Neural Networks for Traffic Flow Prediction

“…Deep learning methods are free from stationary assumptions and effective to capture complex non-linearity using models such as recurrent neural networks (Li et al 2018) and temporal convolutional networks (Wu et al 2019;Li and Zhu 2021). Because traffic data is spatial correlated, CNN (Tang et al 2020) and its extension to arbitrary graphs (Song et al 2020;Tian and Chan 2021) are utilized to capture spatial correlations. Although temporal dependencies and spatial correlations have been considered in these methods, the lack of physical knowledge leads to a lack of generalization ability to out-of-sample scenarios.…”

“…Among the data-driven methods, the most representative branch is traffic flow prediction based on deep learning. For example, using recurrent neural networks (Li et al 2018) or temporal convolution (Wu et al 2019) to model temporal dependencies, using convolutional neural networks (Tang et al 2020) to capture spatial correlations, and using graph convolutional (Song et al 2020;Tian and Chan 2021) to introduce road network information into traffic prediction. In recent years, with a huge volume of traffic data becoming available, the deep learning-based datadriven methods have drawn great attention from both industry and academia, and achieved great success in many real-world applications.…”