Two-Stage Composition of Probabilistic Preferences

Abstract

A strategy of multicriteria decision aid in which initial acceptance of all the alternatives is followed by an automatic classification of them is here proposed. This classification is used in the next step to limit to the alternatives in the upper class the comparisons employed to choose the best alternative. A procedure to perform the classification through successive divisions is proposed. It is also studied the replacement, in the calculation of the preference score according to each criterion, which is a central step in the Composition of Probabilistic Preferences, of the probability of being preferable simultaneously to all alternatives by the average of the probabilities of being preferable to each one. Another development brought is the calculation of preference probabilities based on empirical cumulative distributions, derived from the observed preference counts. Procedures for effectively bringing to practice each of these proposals are presented and the results of applying them to different practical situations are discussed.