Research On Short-term Traffic Flow Combination Prediction Model Based On Temporal And Spatial Correlation

With the continuous development of our economy,the problem of urban traffic congestion is increasingly serious.Accurate traffic flow prediction is very important to alleviate traffic congestion and avoid traffic accidents.Traffic flow prediction is an important link in the construction of intelligent transportation system,which is of great significance to urban traffic planning,traffic management and traffic control.With the comprehensive development of intelligent transportation system,traffic flow prediction has received more and more attention.Traffic itself is periodic,random and uncertain,and the topology structure attribute of road network structure cannot be ignored.Therefore,due to its complex spatio-temporal characteristics,the task of traffic flow prediction is relatively difficult.Complete the following work:(1)Traffic network feature modeling.Based on the graph theory of road network,the complex problems of urban traffic are modeled into a traffic network,and its characteristics are analyzed.The problem of traffic flow prediction can be constructed into a traffic network diagram model,which is generally an undirected graph.The installation location of traffic flow detection sensors on the urban road network is taken as the node,and the sensor location on the adjacent road section is taken as the neighbor node to build the traffic network diagram model.Within the neighborhood,the neighbor node constructs the edge or each edge in the connection diagram reflects the connection between the road sections.(2)A spatio-temporal traffic flow prediction model based on graph wavelet network was proposed in this paper.The external attributes are taken as the attributes of the sections in the road network,and the attributes and traffic characteristics of the sections are modeled to obtain the enhanced feature vector.Secondly,the local and global spatial characteristics of traffic flow are extracted using Graph Wavelet Transform(GWT)and adaptive matrix respectively,and the Gated Recurrent Unit(GRU)is used to extract the time characteristics through the long-term memory ability of the time information.Finally,the attention mechanism is used to capture the temporal and spatial dynamics of traffic flow prediction.Real data sets were used for model prediction,and the research results showed that the prediction effect of A-STIGCN combination model was better than that of traditional linear model and variant model.(3)This paper proposes a spatio-temporal traffic flow prediction model(TSGCN)based on two-flow graph convolutional network.On the basis of A-STIGCN model,the performance was optimized.As dynamic spatial correlation is constrained by fixed road structure and changes in traffic environment,components that affect traffic flow are modeled as dynamic and static components,and the spatial dependence based on road Network topology is learned through Graph Convolution Network(GCN).An adaptive matrix is added to the Gated Linear Unit(GLU)to learn spatial features within a global scope.Meanwhile,the gated linear unit(GLU)is used to process time information efficiently.Finally,the spatiotemporal parts are integrated.Experiments on real data sets show that the prediction accuracy of TSGGN model is better than that of traditional mathematical statistics model,and the prediction error is greatly reduced.At the same time,the model in this paper is better than existing combined models because of the comprehensive consideration of external information affecting traffic flow,and the overall performance of the model has been well verified,so as to provide guidance for traffic control departments.