When handling fuzzy number data, it is common practice to make use of a metric to quantify distances between fuzzy numbers. Several metrics have been suggested in the literature for this purpose. When statistically analyzing fuzzy number-valued data, L 2 metrics become especially useful. This paper introduces a new family of generalized L 2 metrics which take into account key features of the involved fuzzy numbers, namely, a measure of central location and two measures associated with the shape of the fuzzy numbers are used. A crucial property related to these three measures is that necessary and sufficient conditions can be established for them to characterize fuzzy numbers. Furthermore, the family of generalized L 2 metrics depends on one parameter. A discussion is provided regarding the interpretation of this parameter which can guide selection of its value in practice. * corresponding author: magil@uniovi.es, Fax: (+34) 985103354.
Preprint submitted to Fuzzy Sets and SystemsJune 7, 2013
AbstractWhen handling fuzzy number data, it is common practice to make use of a metric to quantify distances between fuzzy numbers. Several metrics have been suggested in the literature for this purpose. When statistically analyzing fuzzy number-valued data, L 2 metrics become especially useful. This paper introduces a new family of generalized L 2 metrics which take into account key features of the involved fuzzy numbers, namely, a measure of central location and two measures associated with the shape of the fuzzy numbers are used. A crucial property related to these three measures is that necessary and sufficient conditions can be established for them to characterize fuzzy numbers. Furthermore, the family of generalized L 2 metrics depends on one parameter. A discussion is provided regarding the interpretation of this parameter which can guide selection of its value in practice.Keywords: fuzzy number, L 2 -metric, wabl/ldev/rdev representation of fuzzy numbers, L 2 wabl/ldev/rdev metric, weighting parameter