The invention provides a traffic flow prediction method and system based on a time-space synchronization dynamic graph attention network, and the method comprises the steps: obtaining traffic input data, carrying out the dynamic feature enhancement processing of the input data through a dynamic feature enhancement module, distributing a dynamic weight for a time interval feature, adjusting the internal correlation weight of the traffic data, and carrying out the prediction of the traffic flow. The spatial-temporal feature data is obtained, a spatial-temporal coupling module is adopted to represent the complex coupling relation of road network nodes in the spatial-temporal dimension and the dynamic internal interaction of spatial-temporal correlation, a mapping function is obtained, and the spatial-temporal coupling module comprises a dynamic graph attention network and a self-adaptive gating time convolution network; and a time position embedding module is constructed, time position information is brought into a time-space relationship, hidden features of a single node are adaptively learned, time position embedding is obtained, time position embedding and a mapping function are fused, and a traffic flow prediction result is obtained. According to the invention, the accuracy and robustness of traffic flow prediction can be effectively improved.
本发明提供一种基于时空同步动态图注意力网络的交通流预测方法及系统,包括:获取交通输入数据,采用动态特征增强模块对输入数据进行动态特征增强处理,为时间间隔特征分配动态权值,调整交通数据的内在相关性权重,得到时空特征数据,采用时空耦合模块表征道路网络节点在时空维度上的复杂耦合关系以及时空相关性的动态内在交互作用,得到映射函数,时空耦合模块包括动态图注意网络和自适应门控时间卷积网络;构建时间位置嵌入模块,将时间位置信息纳入时空关系,并自适应地学习单个节点的隐藏特征,得到时间位置嵌入,融合时间位置嵌入和映射函数,得到交通流量预测结果。本发明能够有效提高交通流量预测的准确性和鲁棒性。
Traffic flow prediction method and system based on time-space synchronization dynamic graph attention network
基于时空同步动态图注意力网络的交通流预测方法及系统
2024-03-08
Patent
Electronic Resource
Chinese
IPC: | G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS / G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen |
European Patent Office | 2023
|Road traffic safety prediction method based on dynamic graph attention space-time network
European Patent Office | 2024
|Traffic flow prediction method and device based on space-time attention graph neural network
European Patent Office | 2024
|Traffic flow prediction method based on space-time attention network
European Patent Office | 2023
|Traffic flow prediction method and device based on time-space synchronization graph aggregation
European Patent Office | 2022
|