Research On Road Traffic State Identification And Prediction Method In Incomplete Information Environment

With the rapid development of modern social economy and the increase of car ownership year by year,the traffic carrying capacity of the road can not meet the increasing traffic demand.Therefore,road traffic congestion not only leads to the reduction of residents ’ travel efficiency,but also brings a series of social and economic problems such as traffic safety and environmental pollution.The development and application of intelligent transportation system has become an effective measure to deal with traffic congestion.Traffic state identification and prediction is an important basis for intelligent transportation systems.However,incomplete information restricts the effect of traffic state identification and prediction.Incomplete information means that the relevant data cannot meet the needs of traffic state discrimination and prediction in one or more aspects such as quality,quantity and coverage.Aiming at the incomplete information environment,this paper mainly studies the data preprocessing of traffic flow parameters,traffic state discrimination based on traffic flow parameters,and short-term traffic flow parameter prediction.The specific research contents are as follows.(1)Selection method of full-state traffic flow parameter data in incomplete information environment.Traffic flow parameter data is an important basis for traffic state identification,and the traffic flow parameter data of some detection points can not cover all traffic states,especially the lack of traffic flow parameter data under congestion.In this study,the k-medoids algorithm is used to cluster the traffic speed data in one day,so as to select the detection points containing the whole traffic state for subsequent traffic state discrimination.Based on the traffic flow parameter data used in this paper,the comparison results show that the accuracy of state discrimination by using the traffic flow parameter data containing the whole traffic state detection points is 90.2 %,and the accuracy of state discrimination by randomly selecting the road sections is only 66.2 %.Therefore,this method can effectively improve the accuracy of traffic state discrimination results.(2)Traffic flow parameter data denoising and prediction method under incomplete information environment.Traffic flow parameter data containing noise components is a type of incomplete information.In this study,a hybrid denoising model based on variational mode decomposition and wavelet threshold(VMD-WT)is proposed.The empirical analysis shows that the proposed denoising method can significantly reduce the influence of noise components on the prediction results compared with the other three denoising methods(WT,EMD-WT,EEMD-WT).In addition,this study uses the long-term and short-term memory network model to predict traffic flow parameters.The empirical analysis shows that the prediction accuracy of the LSTM network model is better than the other three comparison models(ARIMA,GRU,SVM)regardless of whether the data is denoised.(3)Traffic state identification method based on traffic flow parameter data.Based on the traffic flow parameter data,the adaptive spectral clustering algorithm is used to identify the traffic state.The empirical analysis shows that the accuracy of traffic state discrimination results is 90.2 %,which is higher than that of traditional k-means algorithm(accuracy is 88 %),original spectral clustering algorithm(accuracy is 81.1 %)and fuzzy C-means algorithm(accuracy is 88.4 %).Based on the research results,more accurate traffic state prediction results can be obtained through incomplete traffic flow data.It not only provides more accurate decision support for intelligent transportation systems,but also improves the ability of the system to resist incomplete information interference.