Abstract: The econometrics model which is established by traditional ordinary least square mainly reveals the long-term and average relations among economical variables. In the forecasting application, it couldn‘t distinguish near future and far future influence. As a result, its forecasting application can‘ t reveal the short-term waved changes of variables. By using the method that is applied in M-estimation of robust regression to deal with outliers, we set up forecasting robust regression model. Our purpose is to add factors that economical variables are influenced by time sequence into regression model in its forecasting application. By empirical analysis, we also test and verify that this method indeed heighten the model‘s forecasting precision.