On the use of probability and possibility measures in fuzzy clustering

We use a classification model where it is assumed that classes are fuzzy set and the information available is evidential. It has two bad classification loss functions associated, depending on whether plausibility or belief is considered. The plausibility cases of probability and possibility are studied, with a convex combination of losses as an optimality criterion. Both cases lead to linear programming problems whose optimal solutions are optimal classifications, in the case of probability and maximum values of optimal membership functions, for possibility measures