Research On Large-Scale Group Consensus Decision-Making Method Based On Overlapping Communities Under The Influence Of Trust Relationship

With the prosperity and development of society and the advancement of science and technology,the complexity of decision-making problems and decision-making environments has continued to increase.Correspondingly,the scale of decision-makers has continued to expand,and large-scale group decision-making has emerged.Compared with group decision-making problems,large-scale group decision-making problems have many new features.The scale of experts participating in large-group decision-making is large,and dimensionality reduction is required.Decision makers may implement various behaviors that are unfavorable to consensus reaching in the process of consensus decision-making.This makes the traditional group decisionmaking methods unsuitable for solving large-scale group decision-making problems.At the same time,the development of information technology and social media have had a huge impact on the way of communication,driving the development of social network analysis.In large-scale group decision-making,decision-makers are in social networks,and there may be trust relationships among them.The structural characteristics of social network and the trust relationship between decision makers are important factors affecting large-scale group decision-making.Therefore,it is of great theoretical significance and high application value to study the method of large-scale group decision-making influenced by trust relationship.Based on the above background,this thesis studies large-scale group decision-making problem affected by trust relationship.The specific research content and innovation points are as follows:(1)Based on the characteristics of directed networks,a simple and fast overlapping community discovery method is proposed.Firstly,the Speaker Listener Propagation Algorithm is extended to the directed network,and the extended algorithm is further improved based on the in-out degree characteristics of the nodes to improve the performance of the algorithm.The proposed algorithm can adapt to directed unweighted networks and directed weighted networks,and effectively utilizes the directed trust relationship between decision makers.(2)Based on the trust relationship and preference similarity,a comprehensive weight assignment method for subgroup members and subgroup is proposed.For subgroup members,the mixed centrality is used to measure their importance in the subgroup network,and their preference similarity is also considered in weight distribution.For subgroups,trust propagation is used to calculate the trust strength between groups,and the weight distribution of subgroups is carried out by comprehensively considering the trust strength between groups,the degree of consensus inside and outside the group,and the number of members.This method comprehensively considers the influence factors of trust relationship and preference similarity,and can effectively balance the importance of subgroup members and subgroups.(3)The management methods are designed respectively for the non-cooperative behavior,strategic manipulation behavior and withdrawal behavior that decision makers may take during the consensus reaching process.For non-cooperative behaviors,an identification method and weight punishment mechanism based on uninorm operator at the individual level and subgroup level are proposed;for manipulation behaviors,a recognition method based on trust relationship is proposed,and a corresponding weight adjustment mechanism is proposed based on the characteristics of manipulation behaviors;for situations where decision makers withdraw from the consensus-reaching process,a proxy mechanism based on trust relationships is designed based on social network analysis methods.In addition,the behavioral management approaches designed are not conflicting with each other and can be adapted to a variety of decisionmaking situations.Considering the characteristics of the large-scale group consensus decisionmaking problem in the social network environment,this thesis proposes a large-scale group consensus decision-making method based on overlapping community division.The proposed method can be applied to the subgroup division,weight distribution and behavior management problems in actual large-scale group decision-making problems,laying a foundation for further research on this topic.This thesis has 35 pictures,17 tables,and 111 references...