Analysis of selected prioritization methods in the analytic hierarchy process

[摘要] A crucial problem in the analytic hierarchy process (or its extension analytic network process) deriving priorities form pairwise comparison matrices. The most popular methods for deriving priorities are eigenvector method, which is originally proposed by Saaty[5], logarithmic least square method, and least square method. The paper deals with other alternative approaches using goal programming methodology - one of them is based on minimization of sum of absolute or relative deviations and the other one on minimization of maximum deviation. The results of methods are compared on a set of randomly generated matrices of different sizes and consistency levels.

[发布机构] Department of Econometrics, University of Economics, Prague W. Churchill Sq. 4, Praha 3; 13067, Czech Republic^1

[效力级别] 化学

[关键词] Analytic network process;Consistency level;Eigenvector methods;Goal programming;Least square methods;Logarithmic least square methods;Pairwise comparison matrices;Relative deviations