Perova I. Information technology of medical data mining based on hybrid neuro-fuzzy systems.

The thesis is devoted to the solution to the scientific and technological problem of developing and researching information technology to support decision making in the field of medical diagnostics in the online mode in the context of incomplete patient information. Specific features of this problem are the need to process medical information sequentially using Medical Data Mining approaches to support the implementation of eHealth in Ukraine, the need to improve the effectiveness of medical diagnostics in the face of incomplete patient information, and the inability to use traditional Data Mining methods in pure form for processing data streams. Methods for medical data mining during medical online diagnostics in the mode of supervised learning, self-learning and active learning and association based on the hybrid neuro-fuzzy approach have been developed. All developed methods are combined into information technology for medical data mining based on hybrid neuro-fuzzy systems. It consists of three modules: supervised learning, self-learning, active learning and association. Online medical data stream mining based on adaptive neuro-fuzzy systems in the mode of supervised, unsupervised and active learning was considered. Special learning algorithm for neuro-fuzzy systems training was introduced. The proposed information technology allow obtaining additional information about patient diagnosis in conditions of limited a priori information about patient.