PARAMETERISATION OF PROXIMITY FUNCTION IN THE EVALUATION COMPUTATIONAL ALGORITHM USING THE GENETIC ALGORITHM UNDER THE CONDITIONS OF SOURCE DATA UNCERTAINTIES ON THE OBJECTS

The article suggests the new approach for determination of proximity function value, carried out at the second stage by means of evaluation computational algorithm in the field of pattern classification. The issue of comparing the diverse properties of two objects from the specified training set in evaluation computational algorithm is discussed. Comparing the values of the corresponding features under the conditions of source data uncertainties on the objects is considered. The elements of fuzzy sets theory, proximity function in particular, are used to improve the quality of classification in evaluation computational algorithm. The algorithm of using the membership function with two parameters (B and C), determining by means of genetic algorithm is shown.