Sugeno Integral for Rule-Based Ordinal Classification

Quentin Brabant (1) , Miguel Couceiro (2) , Didier Dubois (3, 4) , Henri Prade (3, 4) , Agnès Rico (5)

1 ORPAILLEUR - Knowledge representation, reasonning
2 LAMSADE - Laboratoire d'analyse et modélisation de systèmes pour l'aide à la décision
3 IRIT-ADRIA - Argumentation, Décision, Raisonnement, Incertitude et Apprentissage
4 CNRS - Centre National de la Recherche Scientifique
5 ERIC - Equipe de Recherche en Ingénierie des Connaissances

We present a method for modeling empirical data by a rule set in ordinal classification problems. This method is nonparametric and uses an intermediary model based on Sugeno integral. The accuracy of rule sets thus obtained is competitive with other rule-based classifiers. Special attention is given to the length of rules, i.e., number of conditions.

Domaines: Mathématique discrète [cs.DM]; Apprentissage [cs.LG]

Quentin Brabant, Miguel Couceiro, Didier Dubois, Henri Prade, Agnès Rico. Sugeno Integral for Rule-Based Ordinal Classification. IJCAI Workshop on Learning and Reasoning: Principles and Applications to Everyday Spatial and Temporal Knowledge (IJCAI-ECAI 2018), International Joint Conference on Artificial Intelligence, Jul 2018, Stockholm, Sweden. ⟨hal-01889785⟩