GAM 模型的进一步拓广

Extension of GAM Model

用ＧＡＭ模型模拟预测问题，通常用１－ＡＧＯ生成运算进行数据处理。这种生成运算对波动较小的原始数据预测效果较理想，但对波动大的数据往往失真。此文提出的幂指函数变换法，不受数据波动大小限制，不但拓展了ＧＡＭ模型的应用范围，而且逆变换还原误差也不会扩大。文中用一个实例验证了此种变换的实用性和可靠性。

lmitating prediction problems by GAM model is generally performcd by means of 1-AGO generating calculation in the coarse of data processing. This leincl of generating calculation has better effect on primitive data with less wave movement, and it becomes distortion when the data more wave movement, This paper puts forward transformation approaches of function of Noariables, and prewsents the limits for the magnitude of wave movement of unchanged data. The method not only extednds the application scope of GAM model, but also indicates that reducing error of contravariant trans formation can not be enlarged. The paper testifies that the method is practical and reliable with a practical example.