Traffic Flow Prediction Of Urban Intersections Based On Graph Neural Network

The rapid growth of the urban population brings tremendous pressure to urban traffic.With the deep application of the Internet and the rapid development of artificial intelligence,intelligent transportation has become an essential part of urban construction and development.As a virtual node of the urban road network,accurate traffic flow prediction is significant to an intelligent transportation system.It can help the traffic management department conduct traffic guidance regulations in advance and alleviate traffic congestion.In this paper,the graph convolutional neural network is used to learn the flow prediction problem at urban intersections,and the prototype system is developed.The main work is as follows:Firstly,the original traffic data is preprocessed to obtain the available flow data,and then the intersection flow data is visualized.The traffic network is defined as a directed graph,intersection as nodes,and an edge representing nodes’ connectivity.The traffic network and traffic flow information are modeled as an intersection in flow diagrams and outflow diagrams as the input of the traffic prediction model.An intersection flow prediction model based on GCN-LSTM was proposed to obtain the temporal and spatial dependence of intersection flow using GCN and LSTM.In this model,two components are used to predict the flow,and then the weighted fusion of the two prediction results is carried out to get the final prediction result.Experimental results show that the prediction accuracy based on the GCNLSTM prediction model is good and has a particular application value.On this basis,the Shijiazhuang intersection flow prediction system prototype is developed.The prediction results,traffic conditions,and influencing factors can be displayed through the system.The traffic management department can make corresponding arrangements in advance according to the traffic conditions to alleviate traffic congestion.