Using Choquet integral in Machine learning: What can MCDA bring?
Résumé
In this paper we discuss the Choquet integral model in the realm of Preference Learning, and point out advantages of learning simultaneously partial utility functions and capacities rather than sequentially, i.e., first utility functions and then capacities or vice-versa. Moreover, we present possible interpretation s of the Choquet integral model in Preference Learning based on Shapley values and interaction indices.