Research On Method Of Large-Scale Group Decision-Making With Incomplete Hesitation Fuzzy Information

Large scale group decision-making(LSGDM),as an important branch of modern decisionmaking science,is widely applied in various fields such as enterprise management,quality assessment,resource allocation,etc.The rapid development of information technology makes LSGDM increasingly complex,and it is difficult for decision-makers(DMs)to provide complete and accurate information.As a result,decision information is often incomplete and hesitant.Research on LSGDM with incomplete hesitation fuzzy information has important theoretical significance and practical value.According to the characteristics,this paper mainly conducts research from the three core aspects of supplementing incomplete decision-making information,clustering to reduce the scale,and reaching a consensus.The main work and innovations are as follows:(1)A trust-preference based method for supplementing hesitant fuzzy decision information is proposed to address the issue of incomplete hesitant fuzzy information provided by DMs affecting effective decision-making.Firstly,this method fully considers the asymmetry of trust relationships,trust decay,and the inhibitory effect of distrust relationships when obtaining comprehensive trust.Secondly,we use decision information to define DM preferences,and propose a comprehensive similarity measure for fusing trust and preference to recommend similar neighbors.Finally,a supplementary method is proposed to supplement missing values by using the recommended values of similar neighbors and decision information of the DM.Through experimental verification,the proposed supplementary method not only brings results closer to the original decision information,but also shows good consistency.(2)a bi-objective improved weighted fuzzy clustering method is proposed to address the issues of bias in cluster results caused by neglecting the interest factor,high computational complexity in cluster center calculation,and subjective distance weight settings in existing studies.Firstly,this method defines cognitive similarity and interest similarity,and provides corresponding calculation methods.Based on this,compatibility indicator is proposed to guide clustering.Secondly,the relative standard deviation theory is used to improve the weight calculation,and the compatibility index and weight are used to improve the distance calculation.Finally,the cluster center selection is optimized by the density peak clustering algorithm.It is verified by experiments that the proposed clustering method is not only suitable for different data sets,but also performs well in various indicators and good clustering results.(3)A consensus method based on hybrid dynamic behavior detection and feedback governance is proposed to address the issue of dividing behavior of DMs into cooperative or noncooperative categories,which reduces consensus enthusiasm and efficiency.Firstly,this method recommends modification information based on cluster opinions and its own information.Secondly,cooperation and non-cooperation indices and corresponding calculation formulas are proposed.According to the constituent elements of behavior,modification behavior is divided into three categories: cooperation-guided behavior,non-cooperative-guided behavior,and average behavior.Finally,a personalized governance mechanism is proposed to update the weights of different types of DMs,and at the same time,a floating neutral element is introduced to supervise the group to prevent malicious manipulation.Through experimental verification,the proposed method can effectively improve the consensus efficiency.