Long-time-sequence traffic flow prediction method based on graph convolution-Informer model

Abstract

The invention discloses a long-time-sequence traffic flow prediction method based on a graph convolution-Informer model, and belongs to the technical field of long-time-sequence traffic flow prediction. The technical problem to be solved is to provide an improvement of the long-time-sequence traffic flow prediction method based on the graph convolution-Informer model. According to the technical scheme, the method comprises the following steps that speed information of all passing vehicles at expressway stations and provincial and arterial highway intermodulation stations is collected in unit time, and a traffic flow time sequence information data set is established after data preprocessing; a site network structure topological graph is established according to the relative geographical location information of the expressway stations and the provincial and arterial highway intermodulation stations; a two-layer graph convolutional neural network model structure is constructed, a road network topological structure and traffic flow time sequence information are coded, and spatial dependency feature information of data is learned; coding information obtained through image convolution is input into an Informer layer for training, and data long-time-sequence dependence feature information is learned. The method is applied to traffic flow prediction.

Description

technical field [0001] The invention discloses a long-sequence traffic flow prediction method based on a graph convolution-Informer model, and belongs to the technical field of long-sequence traffic flow prediction methods based on a graph convolution-Informer model. Background technique [0002] With the rapid development of the economy and the rapid growth of the number of motor vehicles, the transportation infrastructure is gradually difficult to keep up with the growing traffic demand. The traffic roads are congested, traffic accidents occur frequently, traffic violations, and illegal phenomena are endless. The inconvenience of people's life has become a It clarifies the important issues that the relevant road traffic management departments need to solve urgently. [0003] On the other hand, with the rapid development of science and technology, especially the breakthroughs in big data processing, cloud computing, data communication, data mining, Internet of Things, artif...

AI Technical Summary

Problems solved by technology

[0008] However, most of the existing model methods are mainly to realize the prediction of traffic flow changes in the short time series in the future. Specifically, in the implementation process of such methods, taking the time interval of 15 minutes as an example, only using the previous hour to three hours (4 to 12 time nodes) time series information trains model parameters for the traffic flow in the next 15 minutes to one hour (1 to 3 time nodes), which leads to the lack of analysis of high-dimensional time series information of traffic characteristics by these methods. Insensitivity to periodic distribution changes of traffic flow in longer time series, insensitivity to traffic flow peaks, etc.

Images