Short-term Traffic Flow Prediction With Residual Graph Attention Network

IR > 计算机与通信学院

	Short-term Traffic Flow Prediction With Residual Graph Attention Network
	Zhang, Xijun; Yu, Guangjie; Shang, Jiyang; Zhang, Baoqi
	2022
发表期刊	Engineering Letters
ISSN	1816-093X
卷号	30 期号:4 页码:1230-1236
摘要	Traffic flow prediction has been essential for traffic management and road network planning. However, the complex urban road network and the strong spatial-temporal correlation of traffic flow data make this problem difficult. Existing prediction methods cannot fully utilize the spatial-temporal correlations in traffic flow data. Therefore, we propose a deep learning model called ResGAT-ABiGRU which combines Residual Network (ResNet), Graph Attention Network (GAT), Attention Mechanism, and the Bidirectional Gated Recurrent Unit (BiGRU). Firstly, GAT is used to capture the spatial correlations of traffic flow data, and then the time characteristics are extracted by Bidirectional GRU. Secondly, the ResNet module stacks multiple GAT layers and designs the attention mechanism to assign weights for different flow sequences to further capture spatial relations. Finally, we obtain the output through the fully connected layers. Validation traffic data from California, USA, is used for verification. The results show that the ResGAT-ABiGRU model proposed in this paper has higher prediction accuracy. Compared the model’s performance with the Gated Recurrent Unit (GRU) baseline model, and the root means square error (RMSE) is reduced by 22.75%, and compared to the T-gcn model, the root mean square error is reduced by 3.29%. © 2022, International Association of Engineers. All rights reserved.
关键词	Deep learning Flow graphs Highway planning Mean square error Roads and streets Street traffic control Attention mechanisms Bidirectional gated recurrent unit Flow data Graph attention neptwork Intelligent transportation Neural-networks Residual neural network Spatial-temporal correlation Traffic flow Traffic flow prediction
收录类别	EI
语种	英语
出版者	International Association of Engineers
EI入藏号	20224813193704
EI主题词	Forecasting
EI分类号	406.2 Roads and Streets ; 432.1 Highway Transportation, General ; 432.4 Highway Traffic Control ; 461.4 Ergonomics and Human Factors Engineering ; 912.2 Management ; 921.4 Combinatorial Mathematics, Includes Graph Theory, Set Theory ; 922.2 Mathematical Statistics
文献类型	期刊论文
条目标识符	https://ir.lut.edu.cn/handle/2XXMBERH/160747
专题	计算机与通信学院
作者单位	Lanzhou University of Technology, Lanzhou; 730050, China
第一作者单位	兰州理工大学
第一作者的第一单位	兰州理工大学
推荐引用方式 GB/T 7714	Zhang, Xijun,Yu, Guangjie,Shang, Jiyang,et al. Short-term Traffic Flow Prediction With Residual Graph Attention Network[J]. Engineering Letters,2022,30(4):1230-1236.
APA	Zhang, Xijun,Yu, Guangjie,Shang, Jiyang,&Zhang, Baoqi.(2022).Short-term Traffic Flow Prediction With Residual Graph Attention Network.Engineering Letters,30(4),1230-1236.
MLA	Zhang, Xijun,et al."Short-term Traffic Flow Prediction With Residual Graph Attention Network".Engineering Letters 30.4(2022):1230-1236.