A Win-Win Multi-issue Service Negotiation Model Based on Bayesian Learning

In this paper we present a bilateral multi-issue negotiation model for service trading which can help negotiators to achieve win-win agreements without disclosing their preferences. A Bayesian technique is used to learn preferences of others and help the offer proposing strategy to generate a win-win offer which is conductive to the achievement of agreements and the long-term customer relationships and profitability. Two improvements are also made to better the effectiveness of the preference learning, one of which is the establishment of two-direction hypothesis spaces for the effectiveness when wrong behaviors of others occur because of wrong hypothesis results at the beginning of the negotiation, and the other one is an updating method of hypothesis probabilities by comparing sequences of issues which can enhance the learning accuracy without having to simulate or learn the conceding strategies of others. Our experimental evaluation shows that the negotiation results of our model are close to or on the Pareto-efficient frontier even when agents make wrong behaviors initially, which means our model can remedy this situation and finally achieve win-win agreements.