Research On Traffic Flow Forecasting For Road Network Based On GAN

The number of vehicles in China has been increasing rapidly year by year.The imbalance between supply and demand of road network resources and vehicles is caused.This problem is embodied in the aggravation of traffic congestion,which has a negative impact on traffic safety and environmental pollution.The intelligent transportation system can improve the cooperation efficiency between pedestrian,vehicles and road nets effectively,so as to improve the operation efficiency of the road network.Traffic flow forecasting of road network,as an important research direction of intelligent traffic system,can provide dynamic and real-time traffic information of road networks.Therefore,the efficient and high-precision road network traffic flow forecasting is of great significance to the construction of an efficient and convenient transportation system.In order to realize an efficient and high-precision road network traffic flow forecasting model,this paper proposes a road network traffic flow forecasting model based on WGAN-GP.The model can generate predicted data similar to the distribution of real road network traffic flow data.This paper firstly preprocesses the traffic volume at the collection point of traffic flow information in the road network into a spatio-temporal traffic volume matrix as the input data of the model according to its time series and spatial geographic information,so that the model can more easily capture the spatio-temporal correlation characteristics of the road network traffic flow.Secondly,a traffic flow prediction model based on SRU-Net is proposed.Residual network is introduced into the model,which can effectively solve the problem of network degradation and reduce the difficulty of model training.In addition,the model can enhance the adaptability to traffic flow data and effectively increase the network depth which enables model to capture the remote spatial correlation between collection points in the road network.Finally,a road network traffic flow forecasting model based on WGAN-GP is proposed.The model uses SRU-Net as the generator,and uses dual discriminator which can distinguish the generated predicted data from two aspects: temporal correlation and spatial correlation,so as to enhance the discriminant ability of the model.The predicted data generated by this model is similar to the distribution of the real road network traffic flow data,which can improve the accuracy of the model for the prediction of large-scale road network traffic flow effectively,and more conformance to requirements of real traffic scenarios.The performance of the prediction model is verified by the actual traffic flow data of highways in England.The experimental results show that the model has smaller prediction error and more stable,and higher prediction accuracy for large scale road network traffic flow prediction.The model can conformance to requirements of low prediction error for most sections of large-scale road network.