Constraints associated with Choquet integrals and other aggregation-free ranking devices

Souhila Kaci (1) , Henri Prade (2)

1 CRIL - Centre de Recherche en Informatique de Lens
2 IRIT-ADRIA - Argumentation, Décision, Raisonnement, Incertitude et Apprentissage

The paper contrasts Choquet integral-based aggregation method with a rank-ordering approach based on the specification of constraints induced by generic principles (or by examples of ranking of particular vectors of attribute values). The latter approach does not use any aggregation operation for evaluating the vectors. It is based on the minimal specificity principle (which amounts to demote vectors as much as they violate constraints) in order to get a complete preorder on the vectors. This approach is compared on an illustrative example to a Choquet integral-based method. We point out the specificity of constraints expressing relative importance that underly Choquet integral.

Domaines: Intelligence artificielle [cs.AI]

Souhila Kaci, Henri Prade. Constraints associated with Choquet integrals and other aggregation-free ranking devices. 12th Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems (IPMU 2008), University of Málaga, Jun 2008, Màlaga, Spain. pp.1344-1351. ⟨hal-00800751⟩