The Compositional Modeling Of Wavelet Analysis And Bayesian Estimation With Statistical Research Of The Railway Freight Volume To Our Country

This paper overcomes the limitations of single forecasting model in railway freight volume forecasting and reduces the error of the model parameters estimated.This paper gives the general scholars to provide a way to improve the model accuracy: the first is from data pretreatment to improve;the second is to change the parameter estimation method;the third is the result of the model.In order to understand the above the three aspects in detail,this work is reflected in:1.Article summarized the research contents,research methods,the research of innovative point and research purpose and meaning.2.The third chapter,by introducing the wavelet analysis to decomposite for our country railway freight volume of non-stationary series,the sequences decomposed establish time series models to forecast.Compared with prediction results of traditional ARMA model,we can be found it is better for the introduction of wavelet analysis of time series model.3.The fourth chapter is based on BIC to determine the order of the model for the ARMA model of sub sets.And it is estimated to solve the bayesian estimation of the ARMA(p,q)model parameters by using winbugs14 software.It gives the convergence test of model parameters and the parameters of the posterior distribution.Through comparing the different parameter estimation of ARMA model prediction results,the conclusion is based on the bayesian estimation of ARMA model to improve the accuracy of the prediction.4.The fifth chapter is based on the research result of the third chapter and the fourth chapter.The wavelet analysis and the bayesian estimation method were combined into a model.Compared with the prediction effect of the third chapter and the fourth chapter,the results show that the combination of the two can improve the precision of the model.5.The sixth chapter is to observe slide moving average parameter of the low frequency which is not very ideal in the fifth chapter.Therefore,the remaining is obtained by which frequency sequence item minus the autoregressive prediction residual term.And then,it estimates the remaining items and restructuring sequence.The research results show that has a positive role for the model prediction effect through this correction method.