Oliynyk A. Multiagent methods of neuro-fuzzy network synthesis

The thesis is devoted to a research and development of multiagent methods of neuro-fuzzy networks synthesis. Object of research - process of neuro-fuzzy networks construction. The purpose is the creation of multiagent methods of neuro-fuzzy networks synthesis for the increasing of their approximate and generalizing abilities. Methods of research are: mathematical statistics, simulated annealing, evolutionary search, multiagent methods, methods of fuzzy logic and the theory of artificial neural networks. The new multiagent methods of swarm intelligence for solving of problems of neuro-fuzzy network synthesis are created: feature selection, cluster analysis, fuzzy rules induction, parametrical and structural-parametrical synthesis of neuro-fuzzy networks. The software based on created multiagent methods of swarm intelligence are developed. The developed software can be used for different problemі solving of neuro-fuzzy network synthesis. The experiments of real-world problem solving are carried out. Based on the experiments effectiveness of proposed method are analyzed. Results of this work are implemented at the industrial enterprises and a higher educational institution.