Clustering and Aggregation of Fuzzy Preference Data: Agreement vs. Information

Abstract

We are considering the application of the data analysis methods in the case of the model of group decision making in which n judges express their opinions as to m options in the form of pairwise preference coefficients d ⁱ _kl ∈ [0,1], where i,i ∈ I = {1,…,n}, denotes a judge and k,l ∈ = {1,…, m} denote options considered. Hence, the coefficients must not necessarily belong to {0,1}, but may also correspond to less definite preferences. It is obvious that the need for allowing such “fuzzy” (or “valued”) preference coefficients to be elicitated arises when the options are perceived through a number of criteria on which they may score in a very different manner, and the elicitation of these scores by every judge for every criterion would be much too cumbersome, if at all possible (considering e.g. the problem of identification of all the—independent—criteria).