| Multi-attribute group decision-making(MAGDM)is an important branch of the decisionmaking theory.Its core is that decision makers(DMs)provide evaluations of alternatives in a set of attributes and then rank these alternatives and select the optimal one.Many decisionmaking problems in economy,engineering,management and life can be modeled as MAGDM problem.Since the decision-making environments become more and more complex,the actual decision-making problem involves many factors.It is difficult for DMs to provide accurate and comprehensive evaluation due to the limitations of DMs’ backgrounds,professional knowledge,experiences and so on.Besides,DMs come from different fields which results in inevitable discrepancies among their opinions.Larger discrepancies will cause some negative effects such as low efficiency and unreliable decision-making results.Therefore,scholars attach great importance to the study of group decision making(GDM)and consensus reaching process(CRP)with linguistic fuzzy information.Such a study has also become a hot topic in recent years.Since DMs are connected which will cause some impacts on GDM,the social network analysis(SNA)can promote the efficiency of GDM and reduce the cost of CRP.Therefore,the study on GDM and CRP with the linguistic fuzzy information based on social networks can enrich and improve the existing decision-making theory,which has important theoretical and practical significance.Based on SNA,this thesis proposes some group decision methods and group consensus methods under the fuzzy linguistic environment.The main work of this thesis is divided into the following five aspects.(1)Probabilistic linguistic MAGDM problem considering individual semantic is studied,which contains the determination of individual semantic information and individual attribute weights of DMs as well as the determination of DMs’ weights.Firstly,based on the preference information on some pair-wise alternatives and decision matrices provided by the DMs,the programming models are constructed to maximize the consistency between the preference information and the decision matrices.By solving the constructed programming models,the individual numerical values of linguistic terms provided by DMs and the individual attribute weights of the DMs are obtained.Based on the preference information on pair-wise alternatives provided by DMs,a new definition of similarity degree is proposed to determine the weights of DMs.By considering the psychological behaviors of DMs,this thesis combines the prospect theory and gained and lost dominance score method to define the individual prospect gained and lost dominance scores of alternatives.Then,a probabilistic linguistic MAGDM method based on the prospect theory and gained and lost dominance score is proposed.(2)CRP and GDM problems under the linguistic intuitionistic fuzzy environment are studied.Firstly,since the consensus threshold in the existing group consensus research is directly given which lacks objectivity,this thesis defines the lowest consensus threshold as a reference to determine the consensus threshold.Considering DMs’ willingness to modify the evaluation information,this thesis proposes a two-stage CRP method in linguistic intuitionistic fuzzy MAGDM.In the first stage,the linguistic intuitionistic fuzzy values(LIFV)with high linguistic uncertainty degree are modified;in the second stage,the elements with high deviation in LIFV are modified.In order to study the factors that will affect the lowest consensus threshold,this thesis designs numerical experiment I.The results show that the more DMs,the lower the lowest consensus threshold.In order to study the performance of the first stage of CRP,this thesis designs numerical experiment II based on the lowest consensus threshold.The results show that the CRP in the first stage can effectively improve the lowest consensus threshold.Finally,this thesis continues to design numerical experiment III and numerical experiment IV to further verify the rationality and effectiveness of the proposed CRP method.(3)Social network based CRP and MAGDM problem with probabilistic linguistic information is studied.According to the evaluation information provided by DMs,the knowledge degrees of the DMs and similarity degrees between DMs are defined successively.DMs are classified into different knowledge levels by utilizing knowledge degrees and adjustment recommendation direction is determined.Similarity degree based social network is constructed with the aid of similarity degrees between DMs.Then,by comprehensively combining adjustment recommendation direction and similarity degree based social network,trust relationship based social network is constructed which represents the connection between the similarity degree and trust relationship between DMs.At the same time,this thesis proposes a two-stage CRP method.The first stage focuses on pair-wise DMs with low similarity degree to eliminate the internal discrepancies in the decision group.The second stage focuses on the DMs with low average similarity degree to improve group consensus level.After the decision group has reached the consensus,this thesis aggregates the individual probabilistic linguistic evaluations of alternatives to obtain the collective probabilistic linguistic evaluations of alternatives and then generates the rankings of alternatives according to the collective evaluations.(4)Ranking consensus and social network based MAGDM with probabilistic linguistic information is studied.Considering that the ultimate goal of GDM is to obtain the rankings of alternatives and generate the optimal alternative,this thesis studies the ranking consensus problem in probabilistic linguistic MAGDM.In order to reflect the group consensus level on the ranking of alternatives,this thesis aggregates the probabilistic linguistic evaluations of alternatives in individual decision matrices to obtain the individual overall evaluations on alternatives and generate the individual rankings of alternatives.Then,according to similarity degrees between the individual rankings of alternatives,the group consensus level is defined.In order to improve group consensus level,this thesis constructs opinion similarity degree based social network and ranking similarity degree based social network,respectively,and proposes a dual-strategies model to improve the group ranking consensus.The first strategy focuses on the situation that the ranking similarity degree between decision makers is low but the opinion similarity degree is high,while the second strategy considers the situation that the ranking similarity degree and opinion similarity between decision makers are both low.The core difference between the two strategies is the different selection of reference DMs and different way of adjusting evaluations.After the group consensus is reached,this thesis then obtains the comprehensive probabilistic linguistic evaluations of alternatives by utilizing the aggregation operator and derives the rankings of alternatives.(5)Social network based probabilistic linguistic large-scale GDM and CRP considering dynamical limited compromise behavior and trust evolution is studied.Firstly,this thesis comprehensively considers the opinion similarity degrees and ranking similarity degrees between DMs to cluster DMs into different subgroups.Then,according to the trust degrees between DMs in the same subgroup and the trust degrees between DMs in different subgroups,the confidence degrees of subgroups and the trust degrees between subgroups are defined.In order to simplify the CRP,this thesis measures the group consensus level according to the decision matrices of subgroups and proposes a CRP method based on trust evolution and dynamic limited compromise behavior.In this method,the limited compromise thresholds are determined by the confidence degrees of subgroups and trust degrees between subgroups.At the end of each round of CRP,the trust degrees between subgroups are updated according to the change in the similarity degrees between subgroups.Then,based on the trust degree and similarity degree,the minimum cost consensus models are constructed to update the evaluations of subgroups and improve the group consensus level.Finally,according to collective decision matrix,evaluations on alternatives with respect to each attribute are aggregated to obtain the collective evaluations on alternatives and the rankings of alternatives are derived to generated the optimal alternatives. |
PDF Full Text Request