Learning of utilitarian decision model through preferences

Abstract

Our goal is to study a decision maker (DM)’s behavioral process that leads to his/her choice. We formalize the notion of a DM who is striving to make the best choice among the various alternatives. Concretely, we develop an approach to learn the complex decision making model of the DM by fitting the recent attitudinal discrete choice models to the real world data. We take the learning information in the form of the exemplary multi-attribute preferences. First experimental results on a set of 12 benchmark datasets suggest that our approach is not only intuitively appealing and interesting from an interpretation point of view but also competitive to state-of-the-art preference learning methods in terms of the prediction accuracy.