Credibility Assessment And Decision-making Mechanism Results Improvement Group Decision Making

Group decision making (GDM) may be in pursuit of completeness of alternatives, may be in pursuit of accuracy of results among specific set of alternatives. But most of GDM need them both. This thesis suppose that the set of alternatives is completeness, so only the accuracy of results is studied, that is only credibility evaluation of results of GDM is concerned. And a new method to develop mechanism of GDM based on credibility evaluation is proposed. The results of GDM are divided into four categories; evaluation method of each is studied. Evaluation method is divided into qualitative evaluation method and quantitative evaluation method. The basis of credibility evaluation is bias in decision making. The bias is also important basis of method to develop mechanism of GDM. So credibility evaluation of results can promote the improvement of decision making mechanism, while the improvement of decision making mechanism can reduce the bias of results.Decision rule and alternatives are main basis of credibility evaluation. There are also subjective factors and random factors can influence the credibility evaluation. The thesis gives credibility evaluation of results of GDM based on the analysis of results given by experts. The evaluation methods of four categories are as follow.Certainty decision results based on independent assessment of alternatives, the evaluation value of alternative that expert gives obeys normal distribution on the assumption that evaluation value is only influenced by random factor. The evaluation value is transformed appropriately, the transformed evaluation value obeys standard normal distribution. Then Chi-square statistic F statistic and variance analysis are used to make statistical analysis of results of GDM. Finally, overall credibility evaluation of results of GDM is given.Certain results of DM based on comparison evaluation are divided into rank-based results and matrix-based results. For rank-based results, qualitative evaluation and quantitative evaluation of credibility of results are obtained by t statistic and information entropy, respectively. For matrix-based results, credibility evaluation of experts and alternatives are obtained by t statistic based on three-element consistency.Fuzzy results of DM based on evaluation independently, that is the evaluation value of alternative is multi-valued or an interval. The credibility evaluation of fuzzy results is given based on ambiguity and consistency. The ambiguity is used to measure the degree of vagueness of results. The consistency is used to measure the degree of agreement among experts. Credibility evaluation of group is synthesis of credibility evaluation of experts.Fuzzy results of DM based on comparison evaluation are divided into fuzzy rank-based results and fuzzy matrix-based results. For fuzzy rank-based results, information entropy is employed to measure ambiguity and consistency of results. Finally the credibility evaluation result is presented. For fuzzy matrix-based results, the credibility evaluation is obtained according to ambiguity and consistency of results. Credibility evaluation of group is synthesis of credibility evaluation of experts.A new method to improve group-decision-making mechanism is proposed. The group-decision-making process is taken as a dynamic process, during which Nash equilibrium and sub game-perfect Nash equilibrium are obtained. In the thesis, input-bias function and bias penalty function are proposed. By equilibrium analysis of three typical game models, a result, that incentive compatibility constraint can be realized in game model through penalty function, is obtained. Sponsor can change equilibrium point through selection of threshold of deviations, so as to improve the efficiency and credibility of GDM.