Short-term Traffic Flow Prediction Method For Expressway Based On Projection Pursuit

On the one hand,the short-term traffic flow prediction of expressway can provide real-time traffic information for travelers to avoid the blindness of travel,and on the other hand,it can provide decision-making and management basis for the traffic management department.However,it is difficult to predict the short-term traffic flow accurately because of its strong randomness and non-linearity and many influencing factors.Projection Pursuit(PP)is a new statistical method for processing and analyzing highdimensional data.It has been widely used in hydrology,agriculture and other fields and has a good prediction effect.Therefore,based on the expressway network toll data,this paper discusses the short-term traffic flow prediction problem based on projection pursuit,and then takes two parameters of traffic volume and travel time as the research objects to verify the applicability and accuracy of the proposed method.Specifically,the main research work of this paper is as follows:(1)The concept and basic characteristics of Projection Pursuit are introduced,and the modeling steps of finding the best projection direction based on genetic algorithm are described.At the same time,the basic structure and modeling process of BP neural network model are introduced,and the possibility of combining BP neural network model with projection pursuit regression model is introduced.(2)In view of the defects of expressway network toll data,this paper expounds the relevant data pretreatment methods.For redundant data,delete the data directly.For the missing data,the key fields of traffic volume and travel time were used as the starting point for identification and processing,and the historical average method and the knearest neighbor method were used to supplement the missing data.For the noise data of travel time,the noise value is deleted based on the standard deviation of two times of the sample mean.(3)A short-term forecasting model of traffic volume and travel time based on Projection Pursuit Regression(PPR)is proposed,and the influencing factors are selected by grey correlation analysis.In the example analysis,the prediction effects of historical average model,BP neural network model and PPR model are compared and analyzed.Through experiments,it is found that PPR model can fit the short-term traffic volume and travel time curve well.Compared with the historical average model and BP neural network model,the mean absolute percentage error(MPAE)of short-term traffic volume prediction is reduced by 16.166% and 5.761% respectively.The mean absolute percentage error decreased by 3.285% and 0.829% in short-term travel time prediction,respectively,which verified the effectiveness of the method proposed in this paper.(4)A short-term prediction model of traffic volume and travel time based on the coupling error back propagation and projection pursuit(BPPP)is proposed.The experimental results show that the BPPP model can improve the prediction accuracy to a certain extent.Compared with PPR model and BP neural network model,the mean absolute percentage error of short-term traffic volume prediction is reduced by 1.796%and 7.557% respectively.the mean absolute percentage error decreased by 0.257% and1.366% in short-term travel time prediction,respectively,which verified the effectiveness of the model coupling.Through the above analysis,the feasibility of the projection pursuit method in the short-term prediction of expressway traffic volume and travel time is verified.This method can be used as a useful supplement to the existing prediction methods and expand the research ideas.