One Kind of Effective Discrete Transformation on Converting the High Growth Series into the Low Growth One

摘要

To the transformation F(a(s)) = a(s)/_(d~s), this paper brings out two kinds of ways to ascertain parameter d, one way, we let d be arithmetic average of class ration in verse, namely d_1= 1/(n-1)((y~((0))(2))/(y~((0))(1)))+(y~((0))(3))/(y~((0))(2))+(y~((0))(4))/(y~((0))(3))+…+(y~((0))(n-1))/(y~((0))(n-2))+ (y~((0))(n))/(y~((0))(n-1))). The other way, we let d be geometric average of class ration inverse, namely d_2= ((y~((0))(n))/(y~((0))(0)))~(1/(n-1)). Through the close theory testifying, we find that the model given by this paper has white exponential law coincidence property, that is to say, if we build the model to the normal index, the result has not model error. At the same time, this paper points out, the request d > 1 is not appropriate to monotone decreasing series; even precision of model will decrease.