Interval Weight Estimation Methods from a Pairwise Comparison Matrix and Their Advantages

Abstract

Many alternative interval weight estimation methods have been proposed to overcome the representation insufficiency of the estimated interval weights of the conventional interval AHP. It has been shown that the improved estimation methods give more appropriate interval weights and they can be used for safer evaluations of alternatives. In this paper, we compare the conventional weight estimation methods in AHP and interval weight estimation methods in interval AHP by the estimation accuracy in ranking alternatives. Numerical experiments are conducted assuming true interval weights. The weights and interval weights are estimated from pairwise comparison matrix randomly generated from the true interval weights. We evaluate to what extent alternative raking is estimated correctly by the interval weight estimation methods under the minimax regret principle and the maximin principle. It is revealed that interval weight estimation is advantageous.