Study of Multiple Criteria Decision-making Methods on Non-dominated Solutions

• Main Authors: Pei-Jen Tiao, 刁培正
• Other Authors: Jei-Zheng Wu
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/50299818932348778817

碩士 === 東吳大學 === 企業管理學系 === 102 === Multi-criteria decision making has various applications in real life. We face Multiple Criteria Decision-Making (MCDM) problems every day and their criteria very often conflict with each other. MCDM methods have been developed to support decision makers to enhance decision quality. MCDM methods use various calculation methods to evaluate the rank of alternatives. Yet, little evidence could support that the best alternative chosen by MCDM method is same with the decision maker’s ideal alternative. On the other hand, decision makers may provide inconsistent preferences due to cognitive biases. For example, the prospect theory advocates that decision makers show different risk attitudes according to different reference points. Therefore, this study aims to examine and compare existing seven seven MCDM methods including TOPSIS, VIKOR, ELECTRE and the piecewise linear prospect theory method (PLP) in terms of effectiveness by using simulation experiements. Controlling variables include number of alternatives, number of criteria, distribution of data set and dominated or non-dominated data set. We also add three different weight combination in these experiments to see how weights affect the MCDM methods. We test four different utility functions. Except the compensatory, non-compensatory and portion of compensatory utility functions, we also use the prospect theory utility function. After that, we compare the MCDM methods’ ranks with decision maker’s ranks by using assumed preference utility functions. The result shows that interactive methods such as PLP and AHP can provide accurate rank to reflect decision makers’ true wishes. Added to this, PLP can use less interactive times than AHP. All of these methods are based on linear utility function, future research might add different utility concept in MCDM methods.