CLASSIFYING THROUGH AN ALGEBRAIC FUZZY STRUCTURE - THE RELEVANCE OF THE ATTRIBUTES

An appropriate fuzzy structure was previously defined to generate appropriate classifications in a universe of discourse in which the characteristics of the elements are defined through attributes whose values are linguistic labels. The way the structure operates depends upon suitable operations on fuzzy attributes represented by fuzzy numbers. However, this classification method suffers from a main disadvantage, i.e., the inherent rigidity deriving from the circumstance that all the attributes share the same weight and, consequently, are taken into account completely disregarding the fact that usually some attributes are more relevant than others. In this article we analyze a suitable weighting rule allowing us to treat in a different way attributes whose relevance is different. The validity of the approach was checked by means of a case study concerning diets for breast-feeding women. (C) 1995 John Wiley & Sons, Inc.