The invention discloses a traffic prediction method based on dynamic graph convolution, and the method comprises the steps: extracting space-time embedding and a dynamic signal through employing a time embedding generator and an MLP layer, and combining the space-time embedding and the dynamic signal to obtain a dynamic graph embedding for generating a dynamic graph. In order to extract the space-time dependency relationship of the traffic signals, a dynamic graph is used to construct a dynamic graph convolution recursion module (DGCRM) based on RNN to extract the space-time characteristics of the traffic signals. In order to identify a main body signal and an abnormal signal in a traffic signal, a residual decomposition mechanism is used for decomposition to obtain the abnormal signal, and DGCRM is used for modeling prediction. According to the traffic prediction method based on the dynamic graph convolution, the method does not depend on any priori knowledge, and the generated dynamic adjacency matrix can mine the dynamic relation between the nodes according to the current time characteristics and the traffic signals; according to the method, the main body signal and the abnormal signal in the traffic signal can be identified and modeled separately to improve the effect.
本发明公开了一种基于动态图卷积的交通预测方法,采用时间embedding生成器和MLP层来提取时空embedding和动态信号,并将两者结合得到动态图embedding用来生成动态图。为了提取交通信号的时空依赖关系,使用动态图构建了一个基于RNN的动态图卷积递归模块(DGCRM)来提取交通信号的时空特征。为了识别交通信号中的主体信号和异常信号,使用残差分解机制来分解得到异常信号并使用DGCRM进行建模预测。本发明采用上述的一种基于动态图卷积的交通预测方法,该方法不依赖任何先验知识,生成的动态邻接矩阵能根据当前时间特征和交通信号来挖掘节点间的动态关系,该方法可以识别交通信号中的主体信号和异常信号并对其分开建模以提升效果。
Traffic prediction method based on dynamic graph convolution
一种基于动态图卷积的交通预测方法
2023-05-16
Patent
Electronic Resource
Chinese
IPC: | G06Q Datenverarbeitungssysteme oder -verfahren, besonders angepasst an verwaltungstechnische, geschäftliche, finanzielle oder betriebswirtschaftliche Zwecke, sowie an geschäftsbezogene Überwachungs- oder Voraussagezwecke , DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS |
Dynamic graph convolution traffic speed prediction method
European Patent Office | 2020
|Dynamic graph convolution traffic flow prediction method based on spatial-temporal characteristics
European Patent Office | 2023
|Double-end graph convolution traffic flow prediction method with graph learning
European Patent Office | 2023
|European Patent Office | 2023
|