Research on Multi-Agent Automatic Negotiation Based on Machine Learning

At present, applying machine learning theory to the multi-agent automatic negotiation system has become the latest research focus in e-commerce area. The paper aimed at using Bayesian rules to update the environmental information in negotiation, using the Q-learning algorithm of reinforcement-learning to generate the negotiation proposals, and established a multi-agent automatic negotiation model based on machine learning, and expanded the traditional Q-learning algorithm, and designed the dynamic Q-learning algorithm based on the current belief and the recent exploring surplus. Furthermore, the convergence of the algorithm has been verified by experiments.