Research On Traffic Data Processing Technology Based On Generative Adversarial Networks

The intelligent transportation system integrates all kinds of advanced technologies and intelligent applications to serve urban traffic management,aiming at improving the efficiency of road use and optimizing the state of urban traffic,thereby improving the quality of urban traffic and bring convenience to urban residents.However,the problem of missing traffic data has plagued researchers in intelligent transportation systems.The missing traffic data reduces its research value and hinders the further application and research of intelligent transportation system.Therefore,how to carry out efficient data recovery and effective use of missing data has become the important research of intelligent transportation system.In view of the problems caused by missing data,this paper applies two new methods for traffic data processing.Firstly,according to the spatio-temporal features of traffic data,this paper proposes a data recovery algorithm based on the 3D convolutional generative adversarial networks to achieve efficient data recovery in various data missing patterns.Secondly,in order to reduce the preprocessing of data in traffic prediction and effectively utilizing the missing traffic data,this paper proposes ambient generative adversarial networks based on 3D convolutional networks to realize the traffic prediction under the case of data missing.The main research contents of this paper are as follows:(1)In order to make full use of the historical traffic data and spatio-temporal features of traffic data,this paper firstly models the traffic data into the 3D spatiotemporal tensor data based on the spatio-temporal features of traffic data.Secondly,a recovery algorithm for missing traffic data based on 3D convolutional generative adversarial networks is proposed by combining generative adversarial networks with3 D convolutional networks.This model uses the historical traffic data training network to fully extract the data features of historical traffic data;and uses 3D convolutional networks to process the spatio-temporal features of traffic data.Finally,in the data recovery stage,in order to enable the generator to make specific missing data,this paper combines the context loss function and the information of the known data to obtains the recovery results.Experimental results show that the model can effectively improve the data recovery accuracy under various missing patterns.(2)In order to reduce the impact of missing data,this paper proposes ambient generative adversarial networks based on 3D convolutional networks for traffic flow prediction under missing data.Aiming at the problem of data missing,this paper introduces the missing measurement module of ambient generative adversarial networks,so that the generator can directly learn the complete data distribution features by using the missing data.Moreover,this paper studies the influence of weekends,holidays and weather factors on the traffic data prediction algorithm,and designs an influencing factor module to extract data features.The outputs of this module are fusion with the outputs of generator to improve the prediction accuracy.The simulation experiment verifies the effectiveness of the proposed traffic prediction algorithm under the missing data by the actual traffic data in Beijing.