Prediction Of Urban Rail Transit Passenger Flow Time Series Based On Wavelet Analysis And Neural Network

With the network operation of urban rail transit,the scale and complexity of the network are increasing rapidly.However,the prediction results and the actual passenger flow are often different in China's urban rail transit passenger flow forecast.Short-term passenger flow forecasting is the basis of operation management and resource allocation,providing effective data support for traffic matching and transport analysis.In the engineering feasibility study stage and engineering design and construction stage short-term passenger flow forecasting has an indispensable importance.So,the accuracy of short-term passenger flow forecast is an important guarantee for the continuous and efficient operation of the rail transit system.After reading a lot of literature,a comprehensive understanding of the domestic and international research on urban rail transit passenger flow forecast situation,and put forward their own forecasting method,that is,the wavelet analysis method to decompose the original passenger data denoising,making the passenger data smooth and smooth;And then the time series method is used to predict the main signal after the impatient.Finally,the noise of the original signal is predicted by the neural network method,and the prediction of the final time series and the prediction of the neural network are obtained.The feasibility of this method is verified by the comparative analysis of the experimental results.Not only in theory to provide a new forecasting methods and ideas,in practice,but also to the urban rail transit operations through more accurate predictions to carry out a series of more positive organizational arrangements.The main work and achievements of this paper are as follows:(1)Summary of urban rail transit passenger flow forecasting methodsThis paper summarizes the methods of urban rail transit passenger flow forecasting by reading a lot of literature,including four-stage method,time series method,neural network method,wavelet analysis method and emerging technology method.In order to make full use of the advantages of various methods,a prediction method based on wavelet analysis denoising as the pilot,time series analysis and prediction is proposed,and the accuracy of neural network is improved.These three methods are based on statistical theory model,nonlinear theoretical model and neural network theory model,can play the best effect of the model.(2)Wavelet analysis denoisingBased on the interpretation of wavelet theory,wavelet transform,multiresolution analysis and wavelet reconstruction are briefly introduced.In order to select the best wavelet basis and its decomposition layer,the noise performance of the wavelet is compared.The threshold denoising method is selected by the comparison denoising method,and the wavelet threshold denoising method is introduced in detail,including denoising principle,denoising principle and threshold selection rule.(3)Time series analysis and predictionThe models of time series analysis include simple regression analysis and trend extrapolation,exponential smoothing,ARIMA and seasonal adjustment.The ARIMA method is selected based on the characteristics of urban rail transit passenger flow data,and the model,principle and application method of ARIMA method are discussed,including stationary analysis,model identification,parameter estimation and residual verification.(4)Neural network analysis and predictionA brief overview of the development of the neural network method is presented,which introduces the single layer sensor,the back propagation BP neural network,the radial basis RBF neural network and the comprehensive optimization of the neural network.Model,the most widely used BP neural network and radial basis RBF neural network are introduced,and the methods of this paper are selected by their own advantages and disadvantages.(5)Comparison of case analysis resultsThrough the design method of this paper,the prediction results are obtained,and compared with different models and methods,it is concluded that the forecasting method is feasible and effective.But also make the comparison the accuracy after removing holidaies,the effect of the amount of data on accuracy.And the accuracy of the prediction between the different scales are also compared.