No label defined (Q4544896)

The standard dissimilarity of crisp sets is measured by means of the corresponding set difference. In the paper, the authors generalize the dissimilarity description to the case of two or more fuzzy sets by means of so-called divergence measures. Their approach is based on the Łukasiewicz t-norm and t-conorm. Basic properties of divergence measures are first studied on singleton universes, when they are simply special real functions called divergence functions. Any divergence function comparing two fuzzy sets is shown to be associated with some transformation of the classical Euclidean distance of values of the corresponding membership functions. In the \(n\)-dimensional case, \(n\geq 3\), the relationship between \((n-1)\)-dimensional aggregation operators and divergence functions is shown. All results are illustrated on several examples comparing cars. Finally, an application of divergence measures to hierarchical cluster analysis is presented.