Research On Complex Multi-Attribute Group Decision Making Method Based On Cluster Analysis

The research on multi-attribute decision making has become more and more mature.Multi-attribute group decision-making is an important part of it and has a wide range of applications.Due to the complexity of things and the differences in thinking of decision makers,some attributes are expressed in the more easy forms,e.g.,interval numbers and linguistic terms.The multi-attribute group decision-making problem with the hybrid attribute values is studied in this thesis.First of all,the theoretical basis of multi-attribute decision making and group decision making is introduced in this thesis.At the same time,it introduces several traditional methods commonly used in multi-attribute decision making and group decision making and analyzes several commonly used methods.It clarifies the core problems that should be solved when dealing with solving large group decision making problems.In addition,the theoretical basis of clustering analysis and several commonly used clustering methods are studied,and the feasibility and credibility of clustering analysis theory in solving complex multi-attribute group decision making are clarified.Secondly,the decision information expressed with interval numbers and linguistic terms is studied,and the two forms of evaluation information are normalized.The interval numbers are normalized into crisp values by calculating the relative distance between the upper and lower boundaries.Linguistic terms are converted into triangular fuzzy numbers and are then converted into crisp values.Then,this thesis improves the k-means clustering method.Two performance measurement indicators,distance between classes and intra-class compactness,are aggregated by F statistic and further used to determine the optimal initial cluster number k.Through the iterative optimization by genetic algorithm,the best classification result is obtained.Finally,the models of determining expert weights within classes,class compactness,class weights and attribute weights are established,and various indicators are applied to integrate the experts' evaluation information within classes.Experts' evaluation information matrix across different classes are aggregated to obtain the overall fusion matrix.The alternatives are ranked based on the experts' evaluation information matrix within classes and the overallfusion matrix.Comparisons are made between the alternatives rankings.The feasibility of the proposed clustering method is verified by an illustration example.