Radyvonenko O. Models and methods of fuzzy clustering under uncertainty conditions

The dissertation work is devoted to development of intellectual data clustering methods, which allow to conduct information classification under the conditions of a priory uncertainty of a form and clusters quantity and to take into account the intuitional picture of data grouping, thereby ensuring of partition efficiency. In the dissertation there are proposed the method of a fuzzy cluster data analysis on the basis of set partition into the equivalence classes by a fuzzy relation and feature space formation model, used in the tasks of cluster images analysis, which uses the chaotic mapping in contrast to existing ones, that substantially enables to reduce cross-correlation dependences between the elements of images. The obtained results are extended for tasks classes, compression and coding of images, formation of approximate fuzzy model of compound object and methods of biomedical data time series processing, which allowed improve data grouping efficiency under the conditions of a priory uncertaintyof a form and clusters quantity. Synthesized models and methods proved their efficiency in creation of biomedical data analysis systems for epidemic data groups separation at influenza and acute respiratory disease tolerant limits calculation, in decision support systems of aviation engines design for approximate fuzzy model formation and in visual data analysis systems for compression and coding of images.