Study On Multi-attributes Group Decision Making Approachs With Uncertain Preference Information Based On Rough Set And Bayes' Theory

Group decision making problem is a classical problem in decision analysis' theory and application. With increasing requirement of science and democracy in real world decision making, research on group decision making is becoming more and more valuable. However, in many decision making context, decision makers often feel it is difficult to describe their preference information precisely for complexity of problem, time press and unfamiliar problem domain. So group decision with uncertain preference has been an appeal field in recent years. According to Steward's classification, there are three kind of uncertainty: rough, fuzzy and stochastic. Fuzzy preference group decision problem has been researched for many years and tons of literature released. But in real life decision problem, such as R&D in new product, credit evaluation, project investment evaluation, uncertainty of rough and stochastic preference information is main information resource. So those two kinds of uncertainty should be researched in group decision making. This thesis focus on it and the main innovations are as follows:(1) For a kind of group decision classification problem with admissible classification error, a method aggregating group preference from multiple decision tables based on variable precision rough set is proposed. By controlling the admissible level of classification error in each table, the ratio of supporting and opposing decision tables to all decision tables, multiple tables are transformed into a group classification pattern table which includes summary information of all decision tables. And then group's classification preference is attained by analyzing the frequence of each pattern supported by decision makers in different classification error level and group agreement ratio. The solution procedure is given and the efficiency of this method is also showed by an example.(2) The main difficulty with application of many existing multiplecriteria group sorting decision aiding methods lies in acquisition of the decision maker's preferential information. Very often, this information has to be given in term of preference model parameters, like importance weights, substitution rates and various thresholds. It is acknowledged, however, that decision makers may prefer to make exemplary decisions than to explain them in terms of special functional model parameters. For this reason, we present a rough set approach to group sorting decisions making in this paper. To deal with decision of multiple decision makers we extend Dominance-based Rough set by introducing specific concepts related to dominance with respect to minimal profiles of evaluations given by multiple decision makers. Our approach differs from existing group decision approaches for we characterize conditions for a consensus attainable by multiple decision makers considered as a whole. Such a perspective permits to handle interactions between decision makers. The usefulness of this approach is illustrated by an example.(3) To a kind of group decision making with stochastic criteria value and weight, an approach combining aggregation of expert's subject probability and choice of stochastic multicriteria alternatives is approved based on Bayesian theory and Monte Carlo simulation. First, a multivariable normal aggregation model is built to get a consensus probability distribution of each criteria's value. By means of Monte Carlo simulation, weights of each criteria is aggregated and a rank probability index is calculated which represent the probability of different rank an alternative might get. Finally, a holistic rank index embodying decision maker's attitude about risk is calculated as well, which can aid decision maker choice the best alternative. The solution procedure is given and the efficiency of this approach is showed by an example.(4) To a kind of large group decision making with uncertain preference, an aggregating approach based on Bayes' theory and Gabbis sampling is proposed. By using multiplicative model judgment matrices with log-normal errors, this approach get group preferences though Bayesian inference. And uncertain preference is handled by Gibbs sampling which is a MCMC method. To the problem of different preference in large group, a group dividing arithmetic is proposed which divides group based on the distance between individual preference and group preference. The solution procedure is given and the efficiency of this approach is showed by example.(5) A real world group decision making case which is sub-project of 11th five years national science and technology support project named "Analysis and research on sustainable development potential of larger metal resource bases" is studied. The method proposed in chapter5 is used to estimate the weight of evaluation index and the method proposed in chapter6 is use to rank the sustainable development potential of five larger metal resource bases. The result shows the flexibility and validity of those methods.