Revised GMDH-type neural network using principal component-regression analysis

In this study, a revised Group Method of Data Handling (GMDH)-type neural network using principal component-regression analysis is proposed and applied to the nonlinear system identification. GMDH-type neural networks can automatically organize neural network architecture by heuristic self-organization method and structural parameters such as the number of layers, the number of neurons in hidden layers and useful input variables are automatically selected so as to minimize the prediction error criterion defined as Akaike´s information criterion (AIC) or Prediction sum of squared (PSS). But, in the heuristic self-organization method, the multicolinearity generates and the network architecture becomes unstable. In this study, the principal component-regression analysis is used for estimating the parameters of the neurons and stable and accurate multi-layered architectures of the GMDH-type neural networks are organized using the heuristic self-organization.