A study on choosing the sample data of grey forecast model

When we establish a forecast model, we often choose the sample data, but the scheme of choosing the proper sample hasn't been solved up now. It is necessary to establish the feasible rules about how to choose the sample data and how much sample should be chosen. Due to there are much data in the economic variables, the rule must be given. The study in the paper shows that the growth rate is a constant in the forecasted GM(1,1) model and it is important to choose the sample data. If the growth rates of the sample data are close to one another, the average absolute errors are less and the sample data can be regarded as the same circle sample data. The data without the circle should be changed to the same circle data by using buffer operates.