Search academic texts

Information × Registration Number 0216U001314, 0115U001974 , R & D reports Title Information technology for multicriteria synthesis of fuzzy knowledge bases popup.stage_title Head Shtovba Serhiy Dmytrovych, Registration Date 05-02-2016 Organization Vinnitsa National Technical University popup.description2 Object of the study is the identification of multifactorial dependencies in engineering, economics, sociology, medicine, sports and other fields. The goal is improving the quality of multifactor identification of dependencies in uncertain conditions by means of creation the information technology for synthesis of fuzzy knowledge bases with taking into account the criteria of accuracy, transparency, compactness, and interpretability. The information technology provides not only the improving the quality of the identification of multifactorial dependencies, but also automating the intellectual activities that leads significantly reducing the complexity and duration of fuzzy knowledge bases design. Research methods are systems approach, mathematical modeling, fuzzy sets theory, theory of identification, optimization techniques and machine learning. The theory of multi-criteria synthesis of fuzzy knowledge bases and its practical implementation in the form of information technology are carried out. We obtained the following scientific results: 1) the formalization of fuzzy knowledge base quality; 2) models for of various fuzzy knowledge bases quality assessment in figures of accuracy, transparency, complexity, and interpretability; 3) the method of formalizing the quality of the synthesis of fuzzy knowledge base taking into account criteria of accuracy, transparency, complexity, and interpretability; 4) structural identification method according to schema "generation of candidate-rules - rules selection - antecedents reduction " with the correct balance of accuracy, complexity and transparency of fuzzy knowledge base; 5) method of interpretability saving of fuzzy knowledge base during the parametric identification based on control the intersection of neighboring fuzzy terms; 6) method of fuzzy parametric identification for discrete dependencies with new learning criteria, taking into account the membership difference only to the main competitors; 7) method of fuzzy parametric identification for discrete dependencies with non-equilibrium different types of errors on the basis of a new learning criterion, which takes into account the cost matrix and fuzzy distance between the desired and actual results of inference. Research results are prepared for usage in the educational process of computer control systems department on Vinnytsia National Technical University and in the company "Liana" for the prediction of natural gas consumption by individual users. Implementation is carried out in the form of a textbook and demo versions of software modules. Product Description popup.authors Галущак Анастасія Володимирівна Грищук Тетяна Вікторівна Жук Олексій Сергійович Клапша Наталія Валеріївна Козачко Олексій Миколайович Мазуренко Віктор Володимирович Обрізан Максим Володимирович Панкевич Ольга Дмитрівна Тилець Роман Олегович Штовба Олена Валеріївна Штовба Сергій Дмитрович popup.nrat_date 2020-04-02 Close