Selection Of Urban Key Roads And Road Traffic Forecast Based On Graph Neural Network

With the continuous development of China's society and economy,urban transportation problems have become increasingly serious.Analyzing the data generated by cities helps us better solve urban transportation problems.This article defines a critical road in a city and installs data acquisition equipment on these roads.By understanding the traffic conditions of key roads,you can understand the traffic conditions of the remaining roads in the city.In this way,only a small number of roads need to be monitored in the city,which can reduce the workload of city data analysis,greatly improve the efficiency of data analysis,and reduce the cost of city construction.This article uses the common traffic flow prediction scenarios in traffic problems to verify the importance of these critical roads.It mainly uses the GPS data of taxis in Changchun City for statistical analysis of urban road traffic flow.The main tasks include the following aspects:First,the parallel processing method is used to perform road matching on taxi GPS data,which greatly improves the matching efficiency.The data of POI points of interest in Changchun,road network data in Changchun,and ground sense coil data in Changchun were crawled.Get road traffic data based on GPS data statistics.Mesh the city.The road matching is performed on the ground-sensing coil data,and the roads with the ground-sensing coil are considered to be the key roads selected through human experience,and these roads are defined as the initial critical roads.Secondly,the road network map data is used to obtain the road's Embedding coding features through the Graph Embedding series of models in the graph neural network method.The roads are clustered using Embedding features and road attributes,and the key roads based on graph neural networks are selected based on the clustering results.Then,road traffic prediction models are mainly divided into two types: full data mode and sparse data mode.Full data mode uses all road history data,and sparse data mode uses key road history data.The feature engineering of the sparse data model is constructed,which mainly includes five parts: road intrinsic attribute features,road network graph relationship features,POI interest point features,Graph Embedding features,and key road flow features.Various machine learning and deep learning methods were tried to construct road traffic prediction models.Finally,the total cost evaluation criteria for key road selection results are defined,which are composed of time cost,data cost and accuracy cost.The time cost is the time required for model training,the data cost is the ratio of the number of key roads to the total number of roads,and the accuracy cost is the accuracy of the road flow prediction model.The experimental results of the total cost of the full data mode and the sparse data mode are compared,and the feasibility of the application of the sparse data mode in the display scene is verified.The key roads of the initialization mode selection and the graph neural network mode selection are compared,and it is verified that the key road selection of the graph neural network mode has a lower total cost,which can reduce or optimize the existing key roads in the existing city,thereby Reach the purpose of reducing urban construction costs.This paper establishes a traffic prediction model that estimates the remaining roads in the city based on the sparse data of key roads.By reducing the number of monitoring equipment to reduce the cost of urban construction,it is also possible to calculate the flow information of roads that did not have monitoring equipment in the past.Because only the road network map data and road attribute data are needed to select key roads,this method can also be used to assist in designing when planning new urban roads,reducing workload and avoiding manual operation errors.