Research On Consensus Reaching Process Of Large-scale Group Decision-making Problems In Social Network Context

The rapid development of information technology and the popularization of so-cial media such as Weibo and WeChat have changed the way people communicate,as well as the environment and characteristics of decision-making problems.Group decision-making problems have shown a trend from small-scale to large-scale,The relationships among the decision makers(DMs)and the structural characteristics of their social network are important factors that affect group decision-making(GDM)process.Making full use of the social network information among DMs can effec-tively enrich the existing decision-making theories.Under the background of big data,the GDM environment and behaviors have become more and more complex.It is of great theoretical value and practical significance to combine the preferences and social network information of DMs to analyze large-scale group decision-making problems(LSGDM)in such complex situations.This article aims to explore LSGDM under social network context and the detailed results of this thesis are given as follows:(1)The consensus reaching process(CRP)is discussed for a class of LSGDM with incomplete preference information under social network context.We propose a novel framework based on social network to manage the CRP for LSGDM faced with incomplete information.In this framework,the large-scale group is first classified into several smaller sub-groups by using a sub-group detection algorithm,based on the social network.Then,we propose an estimating method based on a collaborative filtering algorithm for estimating the missing preference information of the opinion leaders in each sub-group.The two-stage dynamic influence model for handling the CRP in LSGDM begins when the LSGDM is transformed into several smaller sub-group decision processes.In the first stage,a consensus model,based on DeGroot model,is proposed for the CRP within each sub-group.In the second stage,we consider each sub-group as a decision making unit.By focusing on the consensus problem across the sub-groups,we develop a novel opinion-leaders feedback strategy in order to help the sub-groups revise their opinions,working toward consensus.At the same time we prove the effectiveness of the consensus mechanism.(2)The CRP is investigated for a class of LSGDM under dynamic social in-fluence context.We study the propagation of social influence and the evolution of opinions for a group of DMs who communicate and collaborate to make deci-sions.By combining the rule of PageRank and the primacy effect phenomenon in psychology,a social influence propagation model modeling the evolution process of the DMs' social influence is proposed.Afterwards,an opinion evolution model considering the dynamic social influence is proposed to model the opinion formation process.Furthermore,we establish convergence properties of this nonlinear dynam-ical model for the settings of irreducible and reducible influence matrix,where the reducible one is discussed in two scenarios,respectively,the corresponding graph with globally reachable nodes and without globally reachable nodes.(3)Based on the minimum adjustments,the CRP is considered for a class of LSGDM with game behaviors.The interaction behaviors play a core role in the CR-P.Based on the minimum adjustment-based consensus models,we study two types of interactive behaviors in the CRP:1)the interactions between the coordinator and the DMs;2)the interactions among the DMs.We employ a network game to model the interplay among the behaviors of DMs and use Stackelberg game architecture to design an interactive mechanism between the DMs and moderator.We present an optimization model based on these two games and develop a consensus mod-el with maximum linear-quadratic payoffs and minimum adjustment(MPMACM).In the proposed MPMACM,the coordinator provides compensation strategies and feedback suggestions in an effort to guide the DMs to reach the wished consen-sus level with minimum adjustment,while the DMs adjust their opinions with the aim to obtain their maximum payoffs.We then present the equilibrium analysis for the MPMACM.Accordingly,a modified differential evolution algorithm is pre-sented to solve this optimization model.In addition,we verify the effectiveness of the proposed MPMACM by a minimum adjustment-based consensus model for land demolition compensation,and conduct a sensitivity analysis for the relevant param-eters involved in the model.Finally,a comparative analyse with other optimized consensus models is presented to show the advantages of the proposed model.(4)The CRP is addressed for a class of LSGDM with manipulation and non-cooperative behaviors.In LSGDM problems,it is frequent that some DMs exhibit manipulative and non-cooperative behaviors owing to the different interests they might present.Dealing with such large group implies a need for mechanisms to detect DMs' manipulative and non-cooperative behaviors,which might affect the overall efficiency of the CRP.We introduce a novel framework based on social net-work to analyse the manipulative and non-cooperative behaviors in the LSGDM problems.In the developed framework,we first detect and manage the manipula-tive behaviors based on the social network.Afterwards,for the isolated DMs and DMs within a special sub-network,we develop two different approaches to identify and manage non-cooperative behaviors for the consensus model,respectively.At last,two detailed simulation experiments and comparison analysis under different input parameters are presented to demonstrate the efficiency of the novel approaches for coping with manipulative and non-cooperative behaviors.