With the increase of technological and societal demands, more and more decision makers are involved in the process of group decision making, which is called large-scale group decision making (LGDM). For an LGDM problem with linguistic information, it is common that different decision makers tend to provide linguistic assessments defined on multigranular linguistic term sets due to the difference in knowledge and culture background, and that hesitant fuzzy linguistic term sets (HFLTSs) are used by decision makers to model the hesitancy of their assessments. This article proposes first an algorithm to represent a linguistic distribution assessment (LDA) using a hesitant linguistic distribution (HLD). Two other algorithms are then proposed to transform an unbalanced HFLTS into a balanced LDA and to transform a balanced LDA into an unbalanced LDA, respectively. An approach is then proposed to deal with multiattribute LGDM problems with multigranular unbalanced hesitant fuzzy linguistic information based on these algorithms. In the proposed approach, all unbalanced hesitant fuzzy linguistic information is transformed into LDAs defined on a balanced linguistic term set, and then an LDA-based clustering algorithm is devised to cluster decision makers. Based on the clustering result, decision makers' linguistic distribution decision matrices are further fused to obtain collective assessments of alternatives. In order to provide easy-to-understand linguistic results for decision makers, all LDAs of alternatives are represented by HLDs defined on each decision maker's initial linguistic term set. Finally, an example for the selection of subway lines is used to demonstrate the proposed approach.
2 November 2021 in Journal of the Operational Research Society
Leveraging social network trust relationships among experts to reach consensus has become a popular topic in linguistic group decision making (GDM). However, in linguistic contexts, it is commonly accepted that words mean different things for different people, which indicates the necessity of modeling experts’ personalized individual semantics (PISs). Moreover, experts sometimes may show non-cooperative behaviors during the consensus reaching process (CRP) due to their own interests. As a result, this paper focuses on developing a consensus reaching algorithm with non-cooperative behavior management for PIS-based social network GDM problems. First, linguistic preference relations are transformed into fuzzy preference relations by the PIS model, and then social network analysis techniques are used to obtain experts’ weight vector. Afterwards, we propose a feedback adjustment mechanism to improve experts’ adjustment willingness in CPRs, in which the trust relationships and the PISs of experts are utilized to generate adjustment advice for experts. Furthermore, a non-cooperative behavior management mechanism which dynamically adjusts the trust degrees in social network is devised. Followed by this, a numerical example is provided to demonstrate the proposed algorithm. Finally, detailed simulation results are presented to analyze the influence of different parameters on CRPs and illustrate the validity of the proposed algorithm.