A Knowledge Acquisition Method Based on a Multi-Attribute Utility Model

概要

論文の詳細を見る
In solving a design or planning problem, an expert sometimes selects a specific alternative on the basis of multiple criteria. It is difficult to elicit the expert's knowledge concerned with this kind of decision-making and represent it in the form of if-then rules in order to build an expert system. Here, a method for knowledge acquisition and representation is proposed that can overcome this difficulty. In this method, a multi-attribute utility function is employed as a knowledge representation model, which was originally proposed for decision-making with multiple objectives in multi-attribute utility theory. An initial model is obtained by the conventional utility theory technique. This model is modified in order to eliminate conflicts between conclusions arrived at by the model and the expert. Modifications are performed by using an assumption selection module and a control module for focus on the modification process control. In order to improve the efficiency of modification, heuristics are used for assumption selection, backtracking planning, and so on. The proposed method has been applied to knowledge acquisition problems in the construction scheduling do-main, and its validity as a knowledge acquisition technique has been confirmed.
一般社団法人情報処理学会の論文
1992-03-31