摘 要：针对城市路网短时交通速度预测问题，在考虑路网交通状态时空相关性的情况下，提出一种基于图注意力机制的预测方法。该方法利用图注意力网络（GAT）和门控循环单元（GRU）构建了 GAT-GRU 模型，在路网层面对交通状态的时空相关性进行有效地建模，进而预测路网短时交通速度。以城市道路网的交通速度数据为数据源展开数值实验，结果表明 GATGRU 模型的表现均优于对比模型。
关键词：短时交通速度预测；时空相关性；图注意力网络；门控循环单元
DOI:10.19850/j.cnki.2096-4706.2022.22.021
基金项目：国家自然科学基金资助项目（62002101）
中图分类号：TP18；U491 文献标识码：A 文章编号：2096-4706（2022）22-0086-04
Short-Term Traffic Speed Prediction of Urban Road Network Based on Graph Attention Mechanism
YANG Jingyan1, ZHENG Yuqing1, JING Rongrong1, ZHOU Feng1, ZHANG Kunpeng1,2
(1.College of Electrical Engineering, Henan University of Technology, Zhengzhou 450001, China; 2.Department of Automation, Tsinghua University, Beijing 100084, China)
Abstract: Aiming at the problem of short-term traffic speed prediction of urban road network, a prediction method based on the graph attention mechanism is proposed with the consideration of spatiotemporal correlations of road network traffic states. In this method, the GAT-GRU model is proposed by using Graph Attention Network (GAT) and Gated Recurrent Unit (GRU), which carries out modeling effectively for the spatiotemporal correlations of traffic states at the road network level, so as to predict the road network short-time traffic speed. It takes the traffic speed data of urban road network as the data source to conduct numerical experiments, and the results show that the GAT-GRU model performs better than comparison models.
Keywords: short-term traffic speed prediction; spatiotemporal correlation; graph attention network; gated recurrent unit
参考文献：
[1] 张帆 . 基于神经网络的交通流时间序列预测 [J]. 现代信息科技，2020，4（23）：87-89+93.
[2] CAI P，WANG Y，LU G，et al. A spatiotemporal correlative k-nearest neighbor model for short-term traffic multistep forecasting [J]. Transportation Research Part C，2016，62：21-34.
[3] YAO B，CHEN C，CAO Q，et al. Short-Term Traffic Speed Prediction for an Urban Corridor [J].Computer-Aided Civil and Infrastructure Engineering，2017，32（2）：154-169.
[4] MA X L，TAO Z M，WANG Y H，et al. Long short-term memory neural network for traffic speed prediction using remote microwave sensor data [J].Transportation Research Part C，2015，54：187-197.
[5] WANG J，CHEN R，HE Z. Traffic speed prediction for urban transportation network：A path based deep learning approach [J].Transportation Research Part C：Emerging Technologies，2019，100：372-385.
[6] ZHANG K P，ZHENG L，LIU Z J，et al. A deep learning based multitask model for network-wide traffic speed prediction [J]. Neurocomputing，2020，396：438-450.
[7] BAI J D，ZHU J W，SONG Y J，et al.A3T-GCN：Attention temporal graph convolutional network for traffic forecasting [J/OL]. （2020-07-20）.arXiv：2006.11583 [cs.LG].https：//arxiv.org/abs/2006.11583.
作者简介：杨婧琰（2001—），女，汉族，河南焦作人，本科在读，主要研究方向：交通预测；通讯作者：张坤鹏（1987—），男，河南周口人，讲师，硕导，博士后，主要研究方向：智能交通系统、自动驾驶。