Improvement of Interval Weight Estimation in Interval AHP

摘要

Interval AHP was proposed to estimate interval weights from the viewpoint that the vagueness of human judgement causes the inconsistency of a pairwise comparison matrix. Recently, the quality of the original estimation method of interval weights is discussed. The β-relaxation of minimum widths is proposed for a better estimation. However, the quality depends on the selection of parameter β. In this paper, we propose a parameter-free estimation method. It is shown that the proposed method estimates interval weights easily by solving several linear programming problems. Two kinds of experiments are conducted in order to compare the proposed and previous estimation methods: they are from viewpoints of the estimation accuracy and the dominance consistency. By the experiments, the advantages of proposed estimations over the conventional one are demonstrated.