论文部分内容阅读
GMDH (数据处理的分组方法)是对复杂系统的一种启发式自组织的建模方法,是伊凡赫宁柯(Ivakhnenko,1970)借助生物控制论中的自组织的原理用启发式的方法提出的一套建模方法,这种方法的思路完全不同于现有的其他方法,首先,这一方法是启发式的,不是用解析式的方法推导出来的,而是在借鉴模仿生物系统的某些运动过程而提出来的一套建模方法。其次这种建模过程是自组织的,即自动进行变量组合,筛选以及判断是否得到合适的模型。GMDH 以充分地,合理地利用数据,用局部的简单的算法建立整体上复杂的模型为其主要特点。在变量多,数据少,而现有的其牠建模方法很难胜任建模任务的情形,GMDH 却可以得到十分令人满意的结果。因此,近年来 GMDH 受到人们普遍的重视。在1982年 IFAC 第6届“辨识与系统参数估计”学术会议上被列为系统辨识的概念之一。本文将具体讨论 GMDH 方法的基本思路与算法结构。
GMDH (Grouping Method for Data Processing) is a heuristic self-organizing modeling approach to complex systems. It is a heuristic method by Ivakhnenko (1970) using the principle of self-organization in biological control theory Proposed a set of modeling methods, the idea of this method is completely different from other existing methods, first of all, this method is heuristic, not analytic method derived, but in the reference to the biological system A certain set of modeling methods proposed by some movement process. Second, this modeling process is self-organizing, that is, automatic combination of variables, screening and judging whether the appropriate model. GMDH makes full and rational use of the data and uses the simple local algorithm to establish the overall complex model as its main feature. In the case of many variables and few data, and other existing modeling methods are not suitable for modeling tasks, GMDH can achieve very satisfactory results. Therefore, in recent years, GMDH has received universal attention. It was listed as one of the concepts of system identification at the 6th IFAC Conference on Identification and System Parameter Estimation in 1982. This article will discuss in detail the basic idea and algorithm structure of GMDH method.