Research And Design Of Short-term Traffic Flow Forecast Model Based On Deep Learning

Intelligent Transportation System(ITS)is an intelligent comprehensive system that integrates information communication,traffic control,traffic guidance,artificial intelligence and other functions,and is used to solve and manage traffic-related problems.Among them,short-term traffic flow prediction is an important part of the intelligent transportation system,which can provide decision support for urban traffic planning and provide real-time traffic guidance for people to travel.However,the traffic flow has complex spatial-temporal correlation and nonlinear characteristics.How to extract the spatial-temporal characteristics and the true correlation among the nodes in the road network is of great significance for improving the accuracy of traffic flow prediction.In recent years,with the development of artificial intelligence technology,deep learning has become the main realization method of traffic flow prediction,and the existing prediction model based on deep learning needs to calculate all the road network based on the road network structure information when constructing the relationship among nodes.The correlation coefficient among nodes.This approach not only reduces the versatility of the model,but also fails to reflect the true correlation among nodes in the road network.In response to the above problems,this paper studies traffic flow data features,complex network node relationship extraction methods,and deep learning algorithm applications,and builds a general temporal and spatial dynamic correlation short-term traffic flow prediction model and a short-term traffic flow prediction model based on the real relationship between nodes.Time traffic flow prediction model,the main work of this paper is as follows:A general spatial-temporal dynamic correlation short-term traffic flow prediction model is proposed.The model uses graph attention networks to mine the spatial association relationships that are hidden in the traffic flow feature data that dynamically change over time,adjust the graph convolutional network to extract the spatial characteristics of the road network,and use gated linear units to mine the traffic flow.Time characteristics.The simulation results show that the proposed model can achieve higher prediction accuracy and has better versatility.A traffic flow prediction model based on Node-Walk related node search algorithm is proposed.The model uses the Node-Walk algorithm to search for the relevant node set of each node in the urban road network topology map,uses an independent selfattention mechanism to mine the temporal and spatial characteristics of traffic flow,and conducts feature gating and fusion of the temporal and spatial features to obtain the final prediction result.The simulation results show that the proposed model can achieve high prediction accuracy,and provides a search algorithm for obtaining the node set associated with the target node in the urban road network topology.This paper proposes a general model for predicting the overall traffic flow of the urban road network without road network structure information,and a traffic flow prediction model based on the method of searching the related node set of each node in the urban road network topology.The design of short-term traffic flow forecasting models provides new ideas.