The invention discloses a traffic state prediction method under an abnormal event based on multi-subgraph attention, which is realized by constructing an abnormal event subgraph attention network with multi-source information fusion, and specifically comprises the following steps of: establishing an abnormal event set according to traffic event types with relatively high occurrence frequency in a channel; constructing a road section level subgraph attention module by using floating car data and an abnormal event set on the road section; using traffic data and an abnormal event set collected by a fixed detector to construct a cross point level subgraph attention module; inputting an encoded result into a graph neural network model, and respectively learning space and time correlation; and performing short-time traffic state prediction based on the real-time traffic observation data and the abnormal event data. According to the method provided by the invention, by fusing multi-source information, the traffic state change condition after the abnormal event occurs in the channel is deduced and predicted, and a methodology is provided for researching a road network traffic state monitoring technology and improving road traffic transportation safety and emergency guarantee capability.
本发明公开了一种基于多子图注意力的异常事件下交通状态预测方法,通过构建多源信息融合的异常事件子图注意力网络实现,具体包括:根据通道中出现频率较高的交通事件类型组建异常事件集;利用路段上的浮动车数据和异常事件集构建路段级子图注意力模块;利用固定检测器采集的交通数据和异常事件集构建交叉点级子图注意力模块;将编码后的结果输入到图神经网络模型中,分别学习空间和时间相关性;基于实时的交通观测数据和异常事件数据进行短时交通状态预测。本发明的方法通过融合多源信息,对通道中发生异常事件后的交通状态变化情况进行推演预测,对于研究路网交通状态监测技术、提升公路交通运输安全与应急保障能力提供了方法论。
Traffic state prediction method under abnormal event based on multi-subgraph attention
一种基于多子图注意力的异常事件下交通状态预测方法
2024-02-06
Patent
Electronic Resource
Chinese
Traffic flow model simulation method and system and abnormal traffic event prediction method
European Patent Office | 2022
|Traffic Status Prediction and Analysis Based on Mining Frequent Subgraph Patterns
Trans Tech Publications | 2012
|