Non-parametric interval weight estimation methods from a crisp pairwise comparison matrix
Masahiro Inuiguchi, Shigeaki Innan · 2017
In order to express the vagueness of human judgement, methods for interval weight estimation from a crisp pairwise comparison matrix were proposed in Interval AHP. The interval weights estimated by the original method do not reflect well the vagueness of human judgement existing in the given pairwise comparison matrix. Then β-relaxation of minimum widths and γ-relaxation of minimum weighted widths are proposed for better interval weight estimation methods. However, their qualities depend on the selection of parameters β and γ. To overcome this shortcoming, a parameter-free interval weight estimation method has been proposed. In this paper, we further investigate parameter-free interval weight estimation methods and examine their usefulness by numerical experiments. We show that the parameter-free methods have similar performances to β- and γ-relaxation methods with appropriate parameters although their accuracy scores are a little worse.