抄録
We developed the method which recursively identify the multi-attribute utility function based on the observation data which consist of the paired comparison questions and answers. In this method, the weights (w1, w2, …, wn) of additive utility function U(a)=∑ni=1wi·ui(ai) are regarded as random variables and the gaussian sum approximation method is used to estimate the posteriori probability density function of the weights conditioned by the data of comparative judgements. In order to test the applicability of this method, we applied it to the following two examples,
(1) estimation of the value of time,
(2) evaluation of the psychological burden of trip route.
