Fuzzy equalization in the construction of fuzzy sets. (Q5931183)

For a given data set described by means of some probability distribution, any fuzzy model requires a set of granular fuzzy sets (linguistic labels) to be introduced. This paper provides a fuzzy equalization method of constructing such granular fuzzy sets, leading to semantically and experimentally meaningful information granules. Finally, a Ruspini partition of the universe of discour is obtained, formed by triangular (trapezoidal) fuzzy sets, so that each crisp input is linked to at most two fuzzy sets with positive membership value. Note that several technical aspects are not discussed (or only marginally) and that the fuzzy equalization method is discussed for continuous probability distributions only. Hence the paper can be understood as a starting point for the important issue of a sound choice of linguistic labels for practical applications of fuzzy sets theory (the paper itself brings an application of fuzzy equalization in system design).