Abstract:

For the multistage decisionmaking problem that attribute weights and stage weights are completely unknown and attribute information is expressed as different forms, a new decisionmaking method is proposed. Firstly, mixed attribute information is normalized into the form of interval numbers,and on this basis, the attribute weights are obtained by attribute entropy weight intervals and deviation degree of attribute information. Secondly, the decision information in different stages for each decision object is divided into several aggregations by a sequential clustering method. Thirdly, an optimization model which aims minimizing the sum of squares over the decision vector distance in aggregation is proposed, the stage weights over different aggregations are obtained, and the comprehensive stage weights over all decision objects are achieved. Finally, TOPSIS is introduced to rank the decision objects. A numerical example is introduced to illustrate the feasibility and validity of this approach.