In the article the problem of structural and parametric identification for table dependencies is considered. Mathematical formulation is done for problem of building an analytic function in explicit form that best by some criterion extrapolate a given relationship. Evolutionary method of structural identification is developed. It is allow based on a given linearly independent system of basis functions to determine the optimal structure for some criteria such function that corresponds to the mathematical model of the problem. Using the developed evolutionary method allows to specify the following characteristics of the resulting function as parity, periodicity, monotony, range of values and other. Evolutionary parametric identification method is developed. It allows based on a study of character of input parameters to determine the resulting function without the need for complex mathematical transformations and solving systems of nonlinear equations of several variables. The experimental verification of the developed methods for solving applied problems of identification table dependencies using single- and two-factor analysis is done. Advantages of the proposed evolutionary methods over regression models according to the values of mean square error and Fischer ratio are proven. Author Biography Оксана Юріївна Мулеса, Uzhgorod national university, Narodna 3, Uzhgorod, Ukraine, 88000 Candidate of Technical Science, Associate Professor Department of cybernetics and applied mathematics
Alan : Fen Bilimleri ve Matematik
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|