Using the monthly data of China railway passenger traffic volume from 2005 to 2016， we build the multiple seasonal ARIMA model and forecast from January to June of 2017. Through the analysis of Eviews and R software， two models are obtained. The results show that： （1） The prediction error rate under the two types of statistical software is less than 10%， both of them can predict the future value of railway passenger traffic well. （2） Establishing the multiple seasonal ARIMA under the analysis of Eviews software， the analysis thinking is more rigorous and the operation is more complicated， the average of forecast error rate is 4.59% and it has low accuracy； Using R to analyze and model， we can directly use the function about correlation， parameter estimation and prediction in the package to analyze and prodict， the average of forecast error rate is 3.36%， the operation is more simple and the accuracy of prediction is more high. Through this research， ARIMA（2，1，1）×（1，1，1）12 model can be established by using R software. The prediction accuracy of the model is better， which provides some reference value for predicting the future change of railway passenger traffic volume in China.