Preference disaggregation analysis for sorting problems in the context of group decision-making with uncertain and inconsistent preferences

Abstract

Preference disaggregation analysis is effective for inferring a preference model of decision makers from decision examples. Although multiple criteria sorting problems are often handled within the context of group decision-making, it has not been well studied how to estimate a collective sorting model for a group with uncertain and inconsistent preferences through preference disaggregation. To this end, this paper proposes a threshold-based value-driven preference disaggregation method to solve multiple criteria sorting problems involving multiple decision-makers. We allow decision-makers to reduce the cognitive effort of preference expressions by providing possible classes of reference alternatives or pairwise comparisons between them through uncertain expressions. In addition, we develop a two-stage inconsistency identification and elimination procedure that includes dealing with contradictory preference information provided by a single decision maker and managing disagreement among a group of decision makers. A collective preference model consisting of an additive value function and value thresholds is estimated by linear programming. The proposed method is demonstrated by a practical example of financial investment decisions for government-funded projects. The results show that dealing with uncertain and inconsistent preferences can improve the accuracy of preference disaggregation.