Brice Mayag

Séminaire Jeudi 14 Mai 2009, 14h-16h, salle de conférence, Bât. Dumas, 1er étage

"Une méthode interactive d'aide multicritère à la décision basée sur l'intégrale de Choquet 2-additive / An interactive MCDA method based on the 2-additive Choquet integral"

Brice Mayag, Christophe Labreuche (Thales Research and Technology) et Michel Grabisch (Université Paris I)

Abstract

In the context of MultiCriteria Decision Aid, we present the construction of a simple method based on the interview of the decision-maker and allowing to:

• represent faithfully his preferences;
• be interactive with him i.e. the method gives him recommendations to manage his judgements when inconsistencies occur in his preferences;
• take into account the interactions between criteria;
• be understandable by the expert;
• to have a decision-making process faster, thus reducing the of the expertise;
• be autonomous i.e. be able to use the method without intervention of the analyst.

We use as aggreagation function the Choquet integral w.r.t. a 2-additive capacity which generalizes the well known arithmetic mean and take into account the interactions between two criteria. Our goal is to compute the parameters of the 2-additive Choquet integral. These parameters are: the importance index of each criterion and the interaction index of two criteria. Given a decision problem, the method consists to ask to the DM a preferential information on a reference subset of alternatives called a set of binary actions. Then this information is analyzed by using some characterization of the model used.

• If an inconsistency occurs in his preferences, the method generates understandable explanations of the inconsistencies and repair recommendations. Therefore the expert is invited to modify his preferences.
• If there is no inconsistency in his preferences, we compute all the parameters of the model.