Defuzzification of fuzzy intervals (Q1180246)

Studied is the problem of transformation (defuzzification) of a fuzzy set (fuzzy interval) defined in \(\mathbb{R}\) into its numerical representative. Three commonly used methods being investigated in the paper are: mean of maxima (MOM), centre-of area (COA) and fuzzy mean (FM). Consider \(A: \mathbb{R}\to[0,1]\). The MOM method selects an element in \(\mathbb{R}\) at which \(A\) attains maximum; for several maxima the corresponding numerical values are averaged. The COA method determines a numerical value which equally divides the area under the membership function. The numerical value produced by the FM method is taken as \(\int xA(x)\;dx/\int A(x)\;dx\), where the integrals are taken over the support of \(A\). Considering a position of the resulting numerical representative of the membership function with respect to the region of its arguments with the highest membership value, a fuzzy set is classified as regular or irregular. The general properties of these three methods are investigated. It is also shown that the property of regularity is essentially preserved under arithmetic operations.