Research On The Method Of Identifying And Repairing Traffic Flow Error Data Based On Neural Network

In recent years,with the development of new technologies such as big data,cloud computing,Internet of Things and mobile internet,smart transportation construction has become the key to the development of transportation science,and information technology construction plays an important role in the development of smart transportation system.With the development of economy and the increase of vehicle ownership,the traffic behavior of vehicles produces a large amount of traffic data.The quality of these traffic data is the key to the accuracy of information system.At present,the collection of traffic data is mainly achieved by various types of sensors.However,due to some external reasons,such as weather,facility damage or terminal processing errors,the data is missing or errors.It is necessary to identify and repair these traffic data errors in the construction of intelligent traffic information.Based on this,the identification and repair methods of traffic flow error data are studied in this paper.The work of this paper is supported by the national key research and development project "Key Technology Research and Application Demonstration of Large Data Platform for Road Traffic Transportation"(2017YFC0840200).The main work of this paper is as follows:First of all,according to the overall analysis of the issues studied in this paper,the evaluation index of road traffic condition is studied,and the three characteristics of traffic flow,speed and density are selected as the object of follow-up study;the data anomalies are defined;from the data characteristics level,the challenges faced by traffic flow in the process of error data identification and repair are analyzed.Then,aiming at the challenges faced in the process of identifying real-time traffic flow error data,a radial basis neural network method based on self-organizing mapping(SOM-RBF)is proposed to identify traffic flow error data,and the AUC value is used as an evaluation index of the model.Experiments show that this method can quickly and accurately identify error data.Secondly,based on the analysis of traffic flow missing data repair methods,a long-term and short-term memory neural network method based on particle swarm optimization(PSO-LSTM)is proposed to repair traffic flow missing data in order to improve the accuracy of data repair.For the proposed method,root mean square error(RMSE),mean absolute error(MAE),and mean absolute percentage error(MAPE)are selected as the evaluation indexes of the model.The validity of the proposed method is verified with experimental data.The experimental results show that the method can effectively improve the accuracy of data repair.Finally,on the basis of the above-mentioned error data recognition and repair methods,a platform architecture for traffic flow real-time error data recognition and repair system is proposed according to the characteristics of traffic flow data,which provides reference for the construction and application of real-time error detection and repair of traffic big data.