基于全局时空图卷积神经网络的城市交通流量预测  

Urban Traffic Flow Prediction Based on Global Spatiotemporal Graph Convolutional Neural Network

作　　者：王佳昊[1] 黎文斌 郭仕尧 向平 WANG Jiahao;LI Wenbin;GUO Shiyao;XIANG Ping(School of Information and Software Engineering,University of Electronic Science and Technology,Chengdu 610051,China)

机构地区：[1]电子科技大学信息与软件工程学院,成都610051

出　　处：《计算机科学》2024年第S02期534-542,共9页Computer Science

基　　金：电子科技大学-智小金-智能家居联合研究中心项目(H04W210180);内江市科技孵化和成果转化专项资金(2021KJFH004);四川省科技支撑计划项目(2022YFG0212,2021YFG0024);泸州市科技计划项目(2022-XDY-192)。

摘　　要：交通流量预测在智能交通系统(ITS)中发挥着重要作用,将城市中复杂的时空相关性高效且全面地提取出来是交通流量预测中面临的关键挑战。交通速度不仅在时间维度上具有短期和长期周期性依赖关系,而且在空间维度上具有局部和全局依赖性,现有方法对捕获交通数据的时空依赖关系有一定的局限。为此,文中提出了一种基于全局时空图卷积神经网络(Global Spatial-Temporal Graph Convolutional Network,GSTGCN)的深度学习模型,用于解决在城市交通速度预测的局限性。该模型中存在3种时空分量,可相应地对交通数据中的近期、天周期、周周期这3种不同的时空相关性进行建模。每个时空分量都由时间模块和空间模块组成,时间模块为了更好地获取交通数据的时间维度信息,引入了Informer机制以自适应地分配特征权重。空间模型为了更好地获取交通数据的空间关系,引入了图卷积神经网络来提取交通数据的局部和全局空间信息。在两个不同的真实数据集上进行了测试,结果表明所提出的GSTGCN优于最先进的基线模型。Traffic flow prediction plays an important role in intelligent transportation systems(ITS).The key challenge in traffic flow prediction is to efficiently and comprehensively extract the complex spatiotemporal correlations in cities.Traffic speed has not only short-term and long-term periodic dependencies in the temporal dimension,but also local and global dependencies in the spatial dimension.Existing methods have certain limitations in capturing the spatiotemporal dependencies of traffic data.To this end,this paper proposes a deep learning model based on the global spatialtemporal graph convolutional network(GSTGCN)to address the limitations of urban traffic speed prediction.There are three spatiotemporal components in the model,which can model the three different spatiotemporal correlations in traffic data,namely,recent,daily,and weekly cycles.Each spatiotemporal component consists of a time module and a spatial module.In order to better obtain the temporal dimension information of traffic data,the time module introduces the Informer mechanism to adaptively assign feature weights.In order to better obtain the spatial relationship of traffic data,the spatial model introduces a graph convolutional neural network to extract local and global spatial information of traffic data.In the experiments,the proposed model is tested on two different real-world datasets.The results show that the proposed GSTGCN outperforms the most advanced baseline models.

关 键 词：交通流量预测 全局时空图卷积网络 时空依赖性 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]