Usage of fuzzy classification of distribution type for small volume samples

Abstract

volume samples are analyzed. The fuzzy identification method (proposed by Y. M. Klikushyn) which permits identification distributions under condition of small samples was selected for distribution identification.The further research of features and development of fuzzy identification in small volume samples was conducted, the scope of fuzzy identification usage was expended and a research on probability of dispensation differentiation was carried out in the work.The reliability of classifying samples using the generation of reference samples was defined in the work. The research of reliability on one sample showed that the probability of changing linguistic code is quite large. It is recommended to average several samples to increase the likelihood of accurate identification. Number of samples can be determined for a given probability.The research of fuzzy classification included the development of finding fuzzy estimates algorithm for samples of different sizes, finding the necessary and sufficient number of estimates and choice of choice principles of used estimates from the possible range. The correlations for fuzzy estimates based on the specified volume were obtained. Usage of these ratios allowed expanding the scope of fuzzy classification and creating a library of linguistic codes which greatly simplifies its procedure. Recommendations on the samples number required for achieving necessary classification probability were made.