A Novel Traffic Flow Forecasting Method Based On RNN-GCN And BRB

The phenomenon of traffic congestion is not only a major problem in domestic traffic research,but also an important issue that countries in the world need to solve urgently.The solutions implemented by various countries to solve traffic congestion indicate that in order to fundamentally solve the problem of traffic congestion,the flow of traffic must be effectively prediction.Existing prediction models usually have the following problems:(1)Most models focus on single traffic flow data,and do not consider multiple traffic flow factors.Due to the complexity and nonlinearity of the traffic scene and the close relationship between traffic flow speed and traffic flow,therefore,only Consider a traffic factor as a data set to predict traffic flow,which reduces the accuracy of its prediction to a certain extent;(2)The current research on traffic flow prediction methods only predicts a certain time period in the future based on the traffic factor at the moment Traffic flow does not take into account the historical time flow of the sum to predict the traffic flow,that is,does not consider the time correlation,which makes the accuracy of traffic flow prediction to a certain extent reduced;and most of the prediction models are only for relative rules The traffic flow data is processed with a regular graph data structure,but due to the irregularity of traffic flow information,the traditional convolution model cannot perform traffic prediction well.In order to solve the above problems,this article has carried out related research:(1)Based on the information fusion method of the Belief Rule Base(BRB)expert system,a new traffic flow prediction model based on BRB and RNN-GCN is proposed.The BRB system is composed of a series of confidence rules.It is essentially an expert system that can effectively use various types of information to establish a nonlinear model between input and output.In this paper,the data of traffic flow and traffic speed are fused,cluster analysis is used to analyze the fusion output,and the distribution characteristics of the output are obtained to determine the relevant parameters of the system,such as the number of rules of the system and the evaluation level of the result,and make full use of the known samples Information,complete the construction of the confidence rule base based on the extraction paradigm,and obtain a new vehicle flow rate data set.(2)On this basis,this paper will fully consider the time correlation of traffic flow information,solve the problem that traditional convolutional neural networks cannot accurately process graph data structures,and propose a recurrent neural network(Recurrent Neural Network,RNN)and Graph Convolutional Network(Graph Convolutional Network,GCN)model to integrate traffic flow prediction model.The model combines RNN and GCN algorithms,and uses RNN to correlate the traffic flow information in a certain period of time with the traffic flow information at the next moment to obtain the correlation of the time series and form a mapping relationship between input data and output data.In this way,when processing the current information,the RNN will consider the information that appeared before.Afterwards,the processed data with time correlation is used for the topology graph structure.GCN can implement convolution operations on the graph structure data.Any data can establish a topological relationship in the normed space and use it to capture the graph structure.Features,effectively extract spatial features from the topology map for learning.This paper verifies the feasibility and accuracy of the model through comparative experiments,and compared with other prediction models,the prediction accuracy of the model has been improved.Finally,for the problem of poor traffic congestion control,this paper designs a prototype of the traffic flow prediction system based on the traffic flow prediction algorithm.The BRB model is used to fuse the data set to obtain a new data set.Carry out feature extraction and model training,and select traffic prediction algorithms,design data parameters,etc.to predict traffic congestion on a certain road in a certain period of time.