The Research On Driven Factors Identification Of Group Consensus Reaching And Consensus Decision Model

In Multi-criteria decision making(MCDM)process,envelope relations between alternatives are almost non-existent.Therefore,it is necessary to make compromise among all criterion during the alternative selection or ranking process.In decision practice,multiple experts are often invited to form decision group to participate in the MCDM through group decision making(GDM)mechanism.The diversity of knowledge structure,culture background and risk attitude are the advantage of the GDM process.However,the diversity of decision styles may also lead to the conflict among individual decision results,which can harm the effectiveness of group decision.In order to enhance the quality of GDM,the GDM model for the consensus reaching(CR-GDM)is introduced to build group consensus based on the individual opinion.In CR-GDM,the consensus driven factor(CDF)is the key factor to induce the individual opinion dynamics and identify the group consensus.So that,the features of CR-GDM should match the CDF.However,there exist some shortcomings in comprehensive CDF identification,adaptive consensus judgement and process-oriented consensus building.This study identifies the representations of three CDFs,including consensus cost,individual utility,and decision knowledge.Accordingly,the feedback mechanism-based CR-GDM,interaction mechanism-based CR-GDM and the learning mechanism-based CR-GDM are developed.The main parts of this study are shown as follow:(1)The feedback mechanism-based CR-GDM driven by the minimizing consensus cost.Firstly,under the hesitant fuzzy preference environment,the consensus cost and consensus degree are measured based on the preference value similarity and alternative ranking similarity,which can depict the consensus state of decision group.Driven by the compatibility with consensus cost CDF,this study develops the feedback mechanism-based CR-GDM,which sets a virtual GDM coordinator to guide the opinion dynamic to reach consensus.The feedback mechanism-based CR-GDM contain four main parts: the hesitant fuzzy preference standardization for maximizing consensus degree,the adaptive consensus threshold estimation based on the cardinal and ordinal consensus feature,the preference modification based for double consensus cost minimizing,and individual opinion aggregation with consensus induced ordered weighted averaging operator.The feedback mechanism-based CR-GDM can prevents group opinion deviating from the individual decision tendency,and control the comprehensive consensus cost effectively.(2)The interaction mechanism-based CR-GDM driven by the maximizing individual utility.Firstly,the positive relation between individual consensus degree and the utility,and the negative relation between consensus cost and the utility are applied to construct the utility function,which satisfies the condition of supermodular game.Considering the internal willing to interact with the similar opinion,this study develops the BA-BSO interaction mechanism-based CR-GDM,which contains the following two stages: BA free scale network-based interaction relation construction.This process reflects the unbalance interaction relations of decision group,and compress the interaction space of extreme individual,preventing the over-spread of extreme opinion;2)The BSO optimization algorithm-based preference interaction mechanism construction.The BSO-based interaction mechanism contains two types of interaction modes,which balance the effectiveness and efficiency of consensus search.And then,the interaction probability is proportional to the personal cooperation willing and the opinion proximity,which can reduce the negative influence of non-cooperator.Finally,the opinion is updated according to the progressive improvement of individual utility,ensuring the effectiveness of interaction process.The CR-GDM based on the interaction mechanism can explain the consensus emergence and construct the adaptive consensus without crisp threshold.The consensus decision result shows ideal robustness with disturbance of the stubborn non-cooperators.(3)The learning mechanism-based CR-GDM driven by the decision knowledge.In decision practice with large scale group,it is hard for decision maker to observe all individual decision preference statistically.Generally,the decision maker can merely observe the local group and estimate the global consensus state.Therefore,the likelihood function of the true group consensus state is identified as the decision knowledge to guide individual to approach the consensus.The learning mechanism-based CR-GDM is developed driven by the decision knowledge,containing the following two parts: 1)Decision knowledge construction based on TOPSIS method.The TOPSIS method is applied to transform the decision performance to the decision state,which contains alternative ordinal information.The decision knowledge likelihood function is estimated with the individual decision state,reflecting the uncertainty of group consensus.2)The individual learning process based on Bayesian rule.The local learning environment is built with complex network for further observation.And then,the Bayesian rule is utilized to design the personal learning mechanism,based on which the private estimation to group consensus is build.With the guidance of private estimation,the decision preference modification rule is developed according to the feature of the TOPSIS.The individual likelihood weight is updated for the convergency of the decision knowledge to ensure the consensus emergence.The learning mechanism-based CR-GDM isn’t sensitive to the network structure.And the decision result is stable with the disturbance stubborn non-cooperator and the traitorous non-cooperator.The learning mechanism-based CR-GDM has ideal robustness.(4)The investment prospect evaluation of industry sector based on AHP CR-GDM research.Since the actively managed funds tend to concentrate on a few industry sectors to make allocation strategy,this research apply the AHP CR-GDM to evaluate the investment prospect of five main industry sector in Chinese stock market.Considering the influence of macro factors on the stock market,the evaluation criteria system contains three first grade criterion: economic factor,policy factor and market factor,which can be decomposed into ten second criterion: economic boom degree,inflation influence,international trade influence,international financial market influence,monetary policy,fiscal policy,industrial policy,political factors,investor sentiment,industry sector valuation.In order to identify the group consensus,the feedback mechanism-based CR-GDM is used to calculate the criterion weight,and the learning mechanism-based CR-GDM is used obtain the group consensus of alternative ranking.The proposed CR-GDM models can match AHP method greatly,showing the effectiveness and efficiency of the proposed models.