自组织方法（GMDH）中准则的抗干扰性

Noise Immunity of Croup Method of Data Handling(GMDH)

自组织方法是近二十年来发展起来的一种新的建模方法，其基本思想源于生物进化和自然选择理论，在运用此方法对复杂系统建模时，一个重要的问题就是在噪声数据情形下，如何找到那些能够筛选出系统正确模型结构的准则，即准则的抗干扰性问题。本文对此进行了一定的探讨，提出了在有限和无限数据样本上准则抗干扰性问题研究的有关理论和方法，对在噪声数据下复杂系统的建模具有指导意义。

CTGMDH is a novel modeling method in the recent years.Its ideas originated from the theory of evolution and natural selection.When modeling by GMDH,the first problem we confronted is how to choose a rational criterion which is able to find the correct model structure of a system under the circumstances of noise data,that is the problem of noise immunity of criteria. In this article,we researched this problem and put forward some theory and methods with respect to this problem,under the circumstances of finite and infinite sample,respectively. The result of our research is benefit to the complex system modeling with noise data.