A P-based Multiple Criteria Ordinal Classification Method Where Classes Are Described by Subsets of “Non-central” Profiles
摘要
A new multiple criteria ordinal classification method is proposed. Classes are described by two disjoint subsets of profiles, which are representative of their least and the most preferred zones. The paper argues in favor of this way to represent the classes. This is particularly important when the elicitation of either “central” or limiting profiles is a hard task for the Decision Maker (DM). This proposal allows us to work with subsets of profiles very close to the boundaries between adjacent classes. An asymmetric preference relation, P, is used to compare actions against those representative subsets composed by “non-central” profiles. Based on a relational system, (P, D), the method fulfills all the paradigmatic consistency properties of multiple criteria sorting methods, including symmetry through the transposition operation. The method is quite general since P is any asymmetric binary preference relation and D is a transitive relation subject to mild requirements. Its main practical importance consists of allowing a better characterization of classes, and consequently, more appropriate assignments.