Research On Application Of Wavelet Neural Network In Passenger Volume Forecasting

Railway transportation plays a very important role in the social life,and it is the backbone of modern transportation.It takes on most of the long distance material transportation and long distance passenger transport in China.China is rich in resources and vast areas,but the layout of industry and the distribution of resources are not balanced.Therefore,it is necessary to make the railway such a large and inexpensive means of transportation,so that all kinds of resources can be evenly distributed so that they can be fully utilized.Secondly,it possesses extensive land and large population in China,and this convenient and safe way of transportation has great advantages,which is incomparable with other means of transportation.In recent years,the speed of high-speed railway construction in China has been increasing faster and faster,and the volume of passenger traffic has been increasing year by year,and the competition in the passenger market is becoming more and more intense.In order to enhance the competitiveness of the railway passenger transport market,we need to arrange all kinds of organization work of railway transportation reasonably to satisfy people's increasing travel demand.To effectively plan and organize passenger transport and to make passengers travel smoothly,it is necessary to accurately analyze and forecast railway passenger flow and provide the basis for the planning and layout of the passenger transport system,which is very important for the decision-makers and managers of the passenger transport.Based on the above background,based on previous studies,this paper establishes a wavelet neural network model to forecast railway passenger traffic volume.Wavelet neural network is a combination of wavelet theory and artificial neural network,which combines the advantages of time-frequency localization of wavelet transform and the self-learning ability of neural network.Therefore,the wavelet neural network has strong ability of approximation and fault tolerance,and has good convergence and robustness,so it can realize the forecastion better.In the wavelet neural network model,using the wavelet basis function as the number of nodes in the hidden layer,the blindness of the BP neural network is avoided,the accuracy of the operation results is more higher,the approximation ability is improved and the convergence rate is faster.In order to avoid the influence of the randomness of the number of nodes of the neural network input layer on the forecastion results,the embedding dimension is obtained by the phase space reconstruction technique,and the embedded dimension is u sed as the number of nodes of the input layer of the wavelet neural network.In addition,in order to improve the forecastion accuracy of the wavelet neural network model,the wavelet de-noising theory is used to denoise the original time series before the forecastion experiment,and the best denoising model is determined by a large number of contrast simulation experiments.At the same time,in order to further explain the influence of wavelet denoising on the forecastion results,the same wavelet neural network model is used to forecast the de-noising and non denoising time series.After the completion of the passenger traffic forecast model,two experiments were carried out in this paper.The passenger volume of China Railway and the railway passenger volume of the United States were forecasted and analyzed.The feasibility of forecasting the passenger volume by the wavelet neural network model was found by the analysis of the results and the error analysis of the two groups of experiments.In addition,the wavelet denoising method can improve the forecastion accuracy of the forecastion model,reduce the forecastion error and make the forecastion result closer to the real value.