Zakhozhay O. Models, Methods and Information Technology of Hybrid Patterns Recognition for the Consolidate Data Processing in Complex Systems

The dissertation researching is dedicated to solving an important scientific and technical problem of ensuring a priory heterogeneous data processing reliability level in complex systems. Wherein the number of processed data comparisons are reducing. It allows reducing the time complexity of the analysis process. This problem is related to the following contradiction. On the one hand, an increasing the number of information signs allows to increase the reliability of the analysis result, but on the other hand, it requires algorithms complication and increasing the time complexity of data analysis and the decision-making process. Analysis and research in the subject area create possibility for complex systems improve the matrix information model and develop the principle of heterogeneous data consolidation for ensure fast and reliable data processing in complex systems with heterogeneous interference and distortion. Methods have been developed for determining the patterns informativity and their selection and ranking for exclude from analysis the data that is uninformative under current external conditions. Also, it allows the constructing a data queue for multithreaded processing in real time. Methods, models, and information technology for hybrid pattern recognition have been developed for make it possible to provide a priory level of complex system state classifying reliability and reducing the amount of data being compared. An information model of decision support was developed to determine the final classification in the case of a multi-alternative analysis result. Applied information technologies were developed and tested for analyze the spatial distribution of the coke cake temperature in coke-chemical production, ultrasonic distance measurement for automation equipment, and to determine the uniqueness of text data. In this case was use of the proposed models and methods allowed to increase the analysis result reliability. The developed models, methods and information technology can be used to organize the of heterogeneous data processing in the complex systems for different applications, if the signs of different nature can be obtained.