Comparing the Ranking Accuracies among Interval Weight Estimation Methods at the Standard, Minimum and Maximum Solutions under Crisp Pairwise Comparison Matrices

Various estimation methods of interval priority weights from a crisp pairwise comparison matrix have been proposed. They have been compared by using the standard solution such that the sum of the centers of interval weights is one, although the estimation problem has generally non-unique solutions. In this paper, the estimation methods are compared by using three representative solutions of the solution set, the standard solution, the minimum solution, and the maximum solution. The accuracy of ranking alternatives is adopted for comparison. More concretely, the orders of alternatives based on the maximin rule and the maximax rule are compared between interval weights estimated from a given pairwise comparison matrix and the assumed interval weights. From the results of the numerical experiment, some new facts are revealed.