Sensitivity analysis method to address user disparities in the analytic hierarchy process. (30th December 2017)

• Record Type: Journal Article
• Authors: Ivanco, Marie; Hou, Gene; Michaeli, Jennifer
• Is Part Of: Expert systems with applications. Volume 90(2017)
• Page Start: 111
• Page End: 126
• Publication Date: 2017-12-30
• Subjects: Analytic hierarchy process -- Sensitivity analysis -- Disparities; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2017.08.003
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: The Analytic Hierarchy Process does not account for user profile disparities. A sensitivity analysis method is proposed to address these disparities. Inputs that most critically impact the aggregated group results are determined. Decision makers are now informed of the most influential elements of the process. Abstract: Decision makers often face complex problems, which can seldom be addressed well without the use of structured analytical models. Mathematical models have been developed to streamline and facilitate decision making activities, and among these, the Analytic Hierarchy Process (AHP) constitutes one of the most utilized multi-criteria decision analysis methods. While AHP has been thoroughly researched and applied, the method still shows limitations in terms of addressing user profile disparities. A novel sensitivity analysis method based on local partial derivatives is presented here to address these limitations. This new methodology informs AHP users of which pairwise comparisons most impact the derived weights and the ranking of alternatives. The method can also be applied to decision processes that require the aggregation of results obtained by several users, as it highlights which individuals most critically impact the aggregated group results while also enabling to focus on inputs that drive the final ordering of alternatives. An aerospace design and engineering example that requires group decision making is presented to demonstrate and validate the proposed methodology.