L. Anastasakis & N. Mort. The development of self-organization techniques in modelling: A review of the Group Method of Data Handling (GMDH). / The University of Sheffield, United Kingdom. - Research Report No. 813, October 2001, 38 P.

Abstract The necessity of modelling is well established since the structural identification of a process is essential in analysis, control and prediction. In the past, limited information on system behaviour has driven researchers to introduce modelling techniques with a broad range of assumptions on systems' characteristics. Statistical modelling methods, which are based on, these assumptions have generally failed to fully capture the dynamic characteristics of the process. The development of neural networks have partly improved the modelling procedure but their high degree of subjectiveness in the definition of some of their parameters as well as the demand of long data samples remain significant obstacles. On the other hand, real world systems like financial markets have a high degree of volatility and the utilisation of long data samples tends to remove and, in effect, filter the dynamic characteristics of the process. The Group Method of Data Handling (GMDH) belongs to the category of inductive selforganisation data driven approaches. It requires small data samples and is able to optimise models' structure objectively. In this report the stages of GMDH development and a broad spectrum of GMDH algorithms will be explored as well as a diversity of applications. A special study on the external criteria - a key feature in GMDH - will be also presented. Finally some key differences between neural networks and GMDH algorithms will be discussed.

Keywords: self-organising modelling, GMDH, data mining, prediction, external criteria

References 176 Items.