Consensus Reaching Mechanism for Classification-Oriented Group Decision Making Under Unpredictable Uncertainty Based on Weighted Average Evidential Fusion Rule

Group decision-making (GDM) is common in practice. When a GDM problem only requires classifying alternatives instead of producing the ranking of alternatives for a preferred choice, it raises an important question about how to design a consensus-reaching process (CRP). Moreover, under uncertain circumstances, a distributed preference relation (DPR) can not only express superior and inferior degrees on discrete linguistic terms, but can also deal with ignorance properly. How to generate DPR based on multiple attributes and unpredictable uncertainty is also an interesting issue. In this paper, the generation of DPRs on multiple attributes considering unpredictable situations is first discussed. Then, the weighted average evidential fusion rule is proposed to aggregate the DPRs on multiple attributes and experts. The proposed evidential fusion rule satisfies all basic properties of an information combination rule. It is also compared with the evidential reasoning algorithm to illustrate its validity. As for the group consensus, the consensus measure is first defined on the classification result of each expert. Consensus identification and adjustment rules are then proposed, where both opinion leader and social network analysis are considered. An illustrative case study is finally presented to demonstrate the effectiveness and advantages of the proposed approach.