Research conclusions ：(1) With the combination forecasting methods, the regression model for the price margin coefficient is built by using the curve fitting method and modified by using the residual series ARMA model and then the regression ARMA combination forecasting model is built to forecast the expected value of the price margin coefficient of railway project for effective control of the project investment. (2) The combination fitting model has good fittingness. By using this model, the material price margin coefficients from 2009 to 2011 were checked and the forecasting values were well fitting with the actual values. (3) The combination forecasting model presented in this paper is available for the investment estimate of the planned project, the investment evaluations of the medium and long term projects, the cost control of the undergoing project and the bidding price of contracting project.