Optimization of composite fuzzy knowledge bases on rules and relations

Abstract

The problem of composite fuzzy knowledge base optimization is considered. In this case, causes and effects are connected by  fuzzy relations, and causes and effects significance measures are connected by classifying fuzzy rules, which can be considered as qualitative solutions of fuzzy relational equations for the given output classes. The problem of composite fuzzy knowledge base optimization amounts to the problem of min-max clustering and consists of selection of such output classes, for which interval solutions of fuzzy relational equations provide for necessary or extremal levels of the inference accuracy and the number of rules. Such an approach allows complexity reduction for the problem of fuzzy knowledge base optimization by consecutive generation and selection of fuzzy relations and rules.

optimal design of fuzzy knowledge bases, min-max clustering, fuzzy relations, solving fuzzy relational equations, classifying fuzzy rules, composite fuzzy rules