Samitova V. Non-numerical data classification and clusterization

The thesis is devoted to research and development of fuzzy clusterization and classification methods for non-numerical data. The fuzzy clustering method for ordinal data without normal distribution based on frequency prototypes and membership functions is proposed. The adaptive fuzzy clustering method for ordinal data using ordinal-numerical mapping is proposed. This method specifically designed to process ordinal data in an online mode. The thesis proposes a number of robust fuzzy clustering methods for ordinal data and the fuzzy clustering methods for categorical data. The double neo-fuzzy neuron for ordinal data and its learning algorithm are proposed. The recommendations on developed methods use in solving practical tasks are proposed.