DI-UMONS : Dépôt institutionnel de l’université de Mons

Recherche transversale
Rechercher
(titres de publication, de périodique et noms de colloque inclus)
2019-06-19 - Colloque/Présentation - communication orale - Anglais - page(s)

Bouyssou D., Marchant T., Pirlot Marc , "Assigning alternatives to the good or bad category based on several limit profiles" in 25th International Conference on Multiple Criteria Decision Making, 112, Istanbul, Turkey, 2019

  • Codes CREF : Intelligence artificielle (DI1180), Modèles mathématiques d'aide à la décision (DI1151)
  • Unités de recherche UMONS : Mathématique et Recherche opérationnelle (F151)
  • Instituts UMONS : Institut de Recherche en Technologies de l’Information et Sciences de l’Informatique (InforTech)
  • Centres UMONS : Centre de Recherche en Technologie de l’Information (CRTI)

Abstract(s) :

(Anglais) Recently, Fernandez et al. (EJOR 2017) have proposed a method for sorting alternatives into ordered categories based on several criteria. This model known as ELECTRE Tri-nB is a variant of ELECTRE Tri. Instead of using a single limit profile to determine whether an alternative reaches at least a certain quality level, it uses several. In this work, we consider a simplified version of ELECTRE TrinB, which lends itself to an axiomatic characterization. Dropping some pecularities of the underlying model, as originally presented, while keeping its essence permits to shed light on the main features of the model and to analyze its properties. More specifically, we characterize the assignments to ordered categories that can be described by the simplified model. This is done in the spirit of previous work by the present authors, in particular the characterization of outranking relations and the noncompensatory sorting (NCS) model (EJOR 2007). In this talk, we shall: - give a flavor of the axioms used to characterize the model; - analyze the model complexity in terms of the number of profiles needed to separate two ordered categories; - tackle algorithmic issues such as checking whether a given ordered partition can be represented in the model and, eventually, eliciting the model's parameters.