Adaptive multi-issue agent-based negotiation model

This paper presents an adaptive agent-based negotiation model to improve the low efficiency caused by the incomplete information and the diversity of backgrounds of the negotiating parties. Our model at first utilizes the Market-Driver Model to make adjustable rates of overall concession, and then learns opponent´s preference using Bayesian learning, at last takes account of the factors affecting the negotiation process (the issue´s weight, the similarity degree of the two sides´ bid, and the available concession range) and allocates the overall concession to each issue appropriately. This model helps maximum the benefit of both sides, improves the efficiency of negotiation. Our experiments show that this model has better adaptability to the changing situation of the environment and the negotiation process. Especially when the preferences of the two sides are quite different, our model can get a better result and create a win-win situation.