A three-level consensus model for large-scale multi-attribute group decision analysis based on distributed preference relations under social network analysis

Large-scale multi-attribute group decision analysis (LS-MAGDA) is common in practical problems. As a type of preference relation, distributed preference relation (DPR) can express the preferred, non-preferred, indifferent, and uncertain degrees of one alternative over another. In LS-MAGDA, conflict between assessment-based clustering analysis and consensus reaching process (CRP) may occur. Different levels of consensus measurement and feedback mechanism are not fully discussed in previous studies. To solve these problems, a trust-confidence analysis (TCA) framework, which takes into consideration both the trust relationship and self-confidence based on social network analysis (SNA), is proposed to let clustering analysis and CRP not influence with each other. Decision makers’ social status and willingness to modify opinions can be reflected in TCA, which facilitates consensus adjustment and reaching process. A consensus measure framework at attribute, alternative and global levels is then proposed. Additionally, consensus feedback mechanism with different identification and direction rules from attribute level to global level is analyzed considering the consensus degree and importance of attributes. The identification rule becomes looser with the increasing of consensus status and decreasing of attribute weights. An illustrative example of product life cycle design is presented to demonstrate the validity and effectiveness of the proposed method in dealing with realistic problems.