| Expressway plays an important role in the modern development and intelligent operation of traffic information.Accessing traffic state information in real time can help managers apply targeted and effective control measures.In order to improve the accuracy of traffic state estimation,the change point recognition method and the statistical prediction method are applied to predict road traffic states in this thesis.This thesis will introduce the following three parts: the selection of traffic parameters,the identification of traffic states,and the prediction of traffic states.(1)Since individuals have different perceptions of traffic states,this thesis introduces the three parameters of traffic flow: velocity,density and flow from the perspective of traffic wave theory.In order to describe the traffic state better,this thesis chooses the velocity parameter as the traffic state prediction parameter.(2)In order to identify the traffic state when the velocity time series suddenly changes,the change point model of the non-stationary velocity time series is studied in this thesis.By establishing the ARIMA(0,1,2)-ARCH(1)model and using the WBS(Wild Binary Segment)algorithm,this thesis gets 14 change points from August 10 th to August 16 th under the SSIC(Strengthened Schwarz Information Criterion).Combined with the actual situation and analyzing the position of the change point can get the traffic state information when the change point occurs.(3)Due to the inevitable noise information in the time series data,this thesis adopts the empirical wavelet noise reduction method to denoise the velocity time series in advance.Based on the denoised velocity data,three combined prediction models(EWT-ARIMA,EWT-NNAR,EWT-NNBR)are constructed.By comparing the average relative errors of the model directly constructed without noise reduction,it’s found that the prediction accuracy of the combined model is improved by 1.86%,0.24% and 1%,respectively.Finally,according to the relationship between the speed threshold and the traffic state discrimination standard,the traffic state information at a certain moment in the future is obtained.In summary,this thesis detects the model change points through the WBS algorithm.The combined models based on EWT(Empirical Wavelet Transform)are presented,the model prediction accuracy is improved and the predicting purpose of the traffic state is achieved. |
