Ідентифікація багатофакторних залежностей за допомогою баз знань з нечіткими правилами різних типів

Abstract

The purpose of the master's qualification work is improving the quality of multivariable dependencies identification by using of knowledge databases with fuzzy rules of different types. The multivariable identification of dependencies with the usage of the knowledge base with fuzzy rules of various types, such as Mamdani, Sugeno, Larsen, Tsukamoto and singleton fuzzy rules is viewed in the work. The format of hybrid fuzzy knowledge database allows us to describe multivariable dependences in different zones of the factor space with fuzzy rules of the most relevant format. An overview of the issues and tasks is conducted in the work. The tasks are: a) to develop a model for fuzzy inference by knowledge database for such 3 cases : 1) Larsen-Mamdani, 2) Sugeno-Singleton-Tsukamoto, 3) When in database are presented rules from first and second cases; b) modify parametric identification method for a hybrid fuzzy knowledge base; c) perform software implementation; d) conduct computational experiments and make economical reasoning. According to the tasks the models of logical output from knowledge base with rules of various types are described, modification of the method of parametric identification the fuzzy knowledge base with rules of different types is conducted. The implementation of multivariable dependence identification by using of the knowledge database with fuzzy rules of different types is executed by using the MATLAB mathematical package and its component - Fuzzy Logic Toolbox. The correctness of the software is confirmed by testing. Also in the work is confirmed the economic feasibility of development.