Abstract
In this paper, a revised GMDH (Group Method of Data Handling) algorithm, which selects optimal partial polynomials in each selection layer, is developed. In the previous GMDH algorithms, the structure of the partial polynomials is fixed by a predetermined description for all possible combinations of two variables, therefore the partial polynomials obtained are not always optimal regression equation. The revised GMDH algorithm in this paper generates in each selection layer an optimal partial polynomial which minimizes the mean square error for the checking date, and polynomials as such are used to construct the multilayered structure. The revised GMDH algorithm is applied to the identification problem of a largespatial pattern of air pollution and the result obtained is compared with that by the basic GMDH algorithm.
