Designing of classification procedures with the use of equality and difference operators

摘要

The paper addresses a problem of classification formulated in a framework of fuzzy relation calculus. An aspect of partial matching is addressed. This enables us to represent an information supporting class membership as well as information of disconfirmatory character. Both of these sources of information are quantitatively characterized by relevant measures known as degrees of equality and inequality (difference), respectively. Two basic forms of fuzzy relational equations are proposed. In sequel a way of aggregating confirmatory and disconfirmatory information characterizing class membership is introduced. Numerical illustration highlights the properties of the method developed.