Intelligent order online negotiation model with incomplete information of opponent

Abstract

For the problem of high delay rate and high failure rate of order online negotiation, based on Zeuthen Strategy, a novel multistage multilateral co-evolution algorithm was proposed. The algorithm applied a new role of agent called coordinator (CO) to effectively control the multilateral negotiation process, and caught the optimal extent of concession with Bayesian Learning by gradually amending the probability distribution of opponent's reserve-price valuation-vector and dynamically adjusting ones own business bid-curve. Combining the synchronous kicking-out mechanism, invalid consultation was avoided. The experimental results show the proposed method sufficiently captures the opponent's information to help agents duly update their belief, and the utility of negotiation behavior is obviously improved.