The invention discloses a traffic prediction method based on a graph neural network, and belongs to the technical field of traffic prediction, and the method comprises the following steps: (1) inputting a traffic graph signal matrix into a time feature extraction block, and capturing time correlation through a wave pressure module, a smoothing module and a gating neural network; (2) inputting the output of the time feature extraction block into a spatial feature extraction block, and extracting spatial correlation through parallel training of adaptive GCN and dynamic GCN; and (3) fusing the output of the time feature extraction block and the output of the space feature extraction block in a fusion and linear processing module, and carrying out linear processing to obtain a final prediction value. The method solves the problems that in an existing method, time feature extraction of traffic data is not sufficient, and most of existing GCNs (Graph Convolutional Networks) use an adjacent matrix based on the actual distance of nodes, so that processing of incomplete and wrong prior road network data is easily affected.
本发明公开了一种基于图神经网络的交通预测方法,属于交通预测技术领域,该方法包括如下步骤:步骤(1)、将交通图信号矩阵输入时间特征提取块,通过波压模块、平滑模块和门控神经网络捕获时间相关性;步骤(2)、将时间特征提取块的输出输入空间特征提取块,通过自适应GCN与动态GCN并行训练提取空间相关性;步骤(3)、在融合和线性处理模块,对时间特征提取块和空间特征提取块的输出进行融合,并进行线性处理得到最终预测值。本发明解决了现有方法中对于交通数据的时间特征提取不充分,以及绝大多数现有的图卷积网络(GCN)模型使用基于节点实际距离的邻接矩阵,在处理不完整和错误的先验路网数据时容易受到影响的问题。
Traffic prediction method based on graph neural network
一种基于图神经网络的交通预测方法
2024-08-02
Patent
Electronic Resource
Chinese
IPC: | G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS |
Traffic flow prediction method based on graph neural network
European Patent Office | 2024
|SDN traffic prediction method based on graph neural network
European Patent Office | 2024
|Traffic Prediction Using Graph Neural Network
IEEE | 2023
|Traffic jam prediction method based on graph neural network
European Patent Office | 2024
|Airport traffic flow prediction method based on graph neural network
European Patent Office | 2023
|