Constructing Belief Functions from Qualitative Expert Opinions

A new method for constructing belief functions from elicited expert opinions is proposed. It consists in representing qualitatively expert opinions in terms of preference relations. These relations are transformed into constraints of an optimization problem whose resolution allows the generation of the least informative belief functions according to some uncertainty measures. Mono-objective and Multiobjective optimization techniques are used to optimize, respectively, one or simultaneously different uncertainty measures.