The invention relates to the technical field of traffic transportation, and discloses a traffic flow prediction method based on feature attention decomposition and a graph convolutional network, which comprises the following steps: decomposing flow data into seasonal terms, trends and remainders, and representing the adaptability of the size of a convolution kernel in the aspect of trend extraction by adopting a multi-head attention mechanism; then, simulating dynamic space-time correlation of the traffic data by using space-time attention; and finally, extracting local space-time dependency of the traffic data by using space-time convolution. Through experimental analysis, the DFAGCN model provided by the invention has the most advanced performance, and the accuracy of traffic prediction can be obviously improved by applying time sequence decomposition to traffic data.
本发明涉及交通运输技术领域,公开了一种基于特征注意力分解和图卷积网络的交通流量预测方法,将流量数据分解为季节项、趋势和余项,采用多头注意力机制来表示卷积核大小在提取趋势方面的适应性;随后,利用时空注意力来模拟交通数据的动态时空相关性;最后,使用时空卷积来提取交通数据的局部时空依赖性。通过实验分析,本发明提出的DFAGCN模型具有最先进的性能,将时间序列分解应用于交通数据可以显着提高交通预测的准确性。
Traffic flow prediction method based on feature attention decomposition and graph convolution network
基于特征注意力分解和图卷积网络的交通流量预测方法
2024-08-09
Patent
Electronic Resource
Chinese
Traffic flow prediction method based on graph attention convolution network
European Patent Office | 2020
|Traffic Flow Prediction Based on Multi-Spatiotemporal Attention Gated Graph Convolution Network
DOAJ | 2022
|Road network traffic flow prediction method based on gated loop graph convolution attention network
European Patent Office | 2023
|