Consensus Reaching with Dynamic Trust Relationships and Cost-Learning in Group Decision Making

In group decision making (GDM), consensus reaching process (CRP) is a very effective tool for decision makers to reach cooperative agreements. Generally, decision makers need to modify their preferences to reach a consensus, and preference-modifications often mean cost, which make the minimum cost consensus models popular. However, in real-world decision problems, the unit costs of decision makers are often uncertain and difficult to obtain. Meanwhile, the trust relationships defined by social network play an important role in social network group decision making (SNGDM). While in practical SNGDM problems, the trust relationships should be related not only to the social network but also to their preferences, and thus the trust relationships should change dynamically with the preference-modifications. Based on the above two points, we develop a CRP with dynamic trust relationships and cost learning (DTCL-CRP), which is more suitable for solving practical decision-making problems. First, a dynamic trust relationships modeling is presented to update the trust relationships among decision makers and eventually update their weights. Besides, a cost-learning mechanism is designed to learn the unit costs of decision makers, which is the basis of the minimum cost consensus model. Next, we conduct a comparative study between the DTCL-CRP and other CPRs, the results show that the DTCL-CRP performs better than other CRPs under different criteria. Finally, we conduct a sensitive analysis to demonstrate the robustness of the DTCL-CRP.