The invention discloses a road network traffic jam prediction method based on multi-weight map three-dimensional convolution. A model used in the method comprises a map three-dimensional convolution module, an accident module and an accident influence module. In the graph three-dimensional module, firstly, the traffic jam spatial relevance under the influence of spatial heterogeneity considering various static external factors (interest points and road structure attributes) is extracted by using weight graph convolution, and then the traffic jam time relevance under the influence of time heterogeneity considering various static external factors is considered by using three-dimensional convolution. According to the accident influence module, a simple neural network is used, and different influences, namely space-time heterogeneity influences, of traffic congestion of accident influences on different time and road sections are extracted according to characteristics (accident types, severity and the like) of an accident, an accident occurrence place and a time length from accident occurrence to a prediction time step length. In order to avoid model overfitting caused by sparse accident data, the accident module and the accident influence module are independently constructed, are independent from the graph three-dimensional module and are used only when an accident occurs on the road network in an adjacent historical time period.
本发明公开了一种基于多权重图三维卷积的路网交通拥堵预测方法,该方法使用的模型包括图三维卷积模块和事故模块事故影响模块。在图三维模块中,首先使用权重图卷积提取考虑多种静态外部因素(兴趣点和道路结构属性)的空间异质性影响下的交通拥堵空间关联性,再使用三维卷积考虑多种静态外部因素的时间异质性影响下的交通拥堵时间关联性。事故模块事故影响模块使用简单神经网络,不仅根据事故本身的特征(事故类型、严重程度等)、事故发生地点和从事故发生到预测时间步长的时间长度,提取了事故影响在不同时间和路段上的交通拥堵的不同影响,即时空异质性影响。为避免稀疏的事故数据造成模型过拟合,事故模块事故影响模块被单独构建,独立与图三维模块之外,且在邻近历史时间段内路网上有事故发生时才被使用。
Road network traffic jam prediction method based on multi-weight map three-dimensional convolution
一种基于多权重图三维卷积的路网交通拥堵预测方法
2023-11-10
Patent
Elektronische Ressource
Chinesisch
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 |
Road section traffic flow prediction method based on time convolution neural network
Europäisches Patentamt | 2020
|Road network traffic flow prediction method based on gated loop graph convolution attention network
Europäisches Patentamt | 2023
|Urban road network traffic flow prediction method based on linear gray convolution model
Europäisches Patentamt | 2021
|Traffic flow prediction method based on adaptive lightweight time convolution network
Europäisches Patentamt | 2024
|Traffic flow prediction method based on global diffusion convolution residual network
Europäisches Patentamt | 2020
|