User Preferences Elicitation in Bilateral Automated Negotiation Using Recursive Least Square Estimation

Abstract

The negotiating agents are trying to reach a quality agreement during the process of automated negotiation. While each agent tries to improve its own utility, the agreement yields when the opponent reach in an acceptable utility as well. Therefore, learning the opponent's preference during the negotiation is a challenging area of research. The opponent preferences modeled by two parameter vectors: the importance of negotiation issues, and the scoring value of each negotiation issue. In this study, the opponent model is updated by using an incremental recursive least square estimator. As time passes, the estimator reaches calculates the more accurate outcomes. By examining different negotiation domains, the computational experiments show the proposed method outperforms the recent studies.