Search

In group decision-making problems, experts often form multiple subgroups depending upon their backgrounds or attitudes. To explore the effect of different types of subgroups' formation on the final decision and to make the decision process more conforming to human custom, we propose a comprehensive study on the effect of "multi-subgroup background" on a basis of weights optimization and Granular Computing techniques. Firstly, a comparison of different types of clustering strategies is realized to find the most suitable method for clustering experts. Since Analytical Hierarchy Process is selected as the fundamental model, it becomes a reciprocal matrices' or eigenvectors' clustering task. Secondly, allocation strategies of information granularity to each expert along with weight to each subgroup are carefully designed and optimized. Information granularity is viewed as a design asset to provide experts with some flexibility to adjust their evaluation results, whereas weights reflect the importance degrees of different subgroups and further help to increase the consensus. A granulation-degranulation process is developed to evaluate the performance of a set of weights and granularities under a multi-criteria objective's guidance. Thirdly, a proper evolutionary optimization method like particle swarm optimization is redesigned to iteratively generate the best set of information granularities and weights under an equality constraint condition. A series of numerical studies on synthetic data and real-world data are executed to verify the effectiveness of our method. The results show same trend present across experts. [ABSTRACT FROM AUTHOR]

Linguistic variables are flexible and intuitive attraction for expressing the wording of decision makers. This paper introduces a new type of linguistic fuzzy sets called linguistic dual hesitant fuzzy sets to express the hesitancy of decision makers' qualitative preferences and non-preferences. Considering the application in decision making, linguistic dual hesitant fuzzy preference relations (LDHFPRs) are introduced that permit the decision makers to apply several linguistic variables to indicate a qualitative preferred judgment and a qualitative non-preferred judgment, respectively. To rank objects from LDHFPRs rationally, a consistency concept is first presented. Then, two optimal models are built to judge the consistency of LDHFPRs. When LDHFPRs are inconsistent, an optimal model-based iteration algorithm for obtaining consistent LDHFPRs is offered. Based on consistent linguistic intuitionistic fuzzy preference relations, a method for calculating the weighted linguistic intuitionistic fuzzy priority vector is introduced. In the setting of group decision making (GDM), a consensus measure based on individually weighted consistent reverse complementary linguistic intuitionistic fuzzy preference relations is defined. When the consensus does not satisfy the requirement, a two-step optimal model-based method for increasing the consensus level is offered. Furthermore, an approach for GDM with LDHFPRs is developed. Finally, an illustrative example concerning the evaluation of basic services internet enterprise websites is provided to show the efficiency of the new method. [ABSTRACT FROM AUTHOR]