A Bayesian approach to portfolios selection in multicriteria group decision making

In the a-posteriori approach to multicriteria decision making the idea is to first find a set of interesting (usually non-dominated) decision alternatives and then let the decision maker select among these.

Often an additional demand is to limit the size of alternatives to a small number of solutions. In this case, it is important to state preferences on sets. In previous work it has been shown that independent normalization of objective functions (using for instance desirability functions) combined with the hypervolume indicator can be used to formulate such set-preferences.

A procedure to compute and to maximize the probability that a set of solutions contains at least one satisfactory solution is established. Moreover, we extend the model to the scenario of multiple decision makers. For this we compute the probability that at least one solution in a given set satisfies all decision makers. First, the information required a-priori from the decision makers is considered. Then, a computational procedure to compute the probability for a single set to contain a solution, which is acceptable to all decision makers, is introduced. Thereafter, we discuss how the computational effort can be reduced and how the measure can be maximized. Practical examples for using this in database queries will be discussed, in order to show how this approach relates to applications.