Nessonova M. Models and methods for patients' degree of severity assessment to support physician's decision

The new decision of topical theoretical and applied task of models and methods development for patients' degree of severity assessment to support physician's decision is suggested in the thesis. The new method for supervised classification is suggested. The method is based on the spatial representation analysis of relationships between classes and describing patterns, which is derived as a result of utilization of methods for geometrical interpretation of multidimensional data structure. Methods for classifiers compositions assembling have been further developed due to the combination of principles of specialization and weighted majority vote. The developed methods to design classifiers by learning information and their compositions were realized in the information technology, and this gives an ability to develop models for patients' degree of severity assessment and for clinical outcome prediction in cases of pancreas trauma and traumatic pancreatitis, bile-excreting ducts diseases, and ischemic and hemorrhagic strokes. The conducted comparative analysis shows that the developed models have higher accuracy in comparison with models, which were developed using classical methods of statistical modeling and analysis.