Multi-attribute decision making: a test on the impact of data attributes dependency

Abstract

[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT AUTHOR'S REQUEST.] In Multi-Attribute Decision Making (MADM) problems, several methods have been developed to aid the decision maker. MADM procedures can be applied to a wide range of human choices, from the professional to the managerial to the political. In reality, SAW and AHP are the two best known and widely used approaches to determine the order or preference of various alternatives. Other methods such as DI, TOPSIS, and WPM are also well-known in MADM research field. Most of the models assume independence among attributes. AHP and DI are two models that claim to work regardless of dependency among attributes, while some claim that dependency does not have significant impact on results. The question though is whether dependency has any impact on decision quality in practice. In this research, we generate data with given levels of dependency and apply selected MADM methods to the data. By comparing the results from these methods, we hope to determine whether dependency among attributes has an impact on the ranking alternatives or not. It should be noted that AHP has a different approach that it does not share the same raw data as other methods so it won't be included in our simulation experiment.