Research On Multi-Agent Negotiation Mechanism Based On RBF Neural Network

The emergence of the electronic commerce is changing the mode of enterprise business operation and people's way of life. Negotiation is the key step of electronic commerce, and using Agent instead of user and businessman for negotiation will improve the efficiency of remote business operation greatly. It is an important problem that how to improve the Agent's learning ability in negotiation process. In theory, it will be helpful for the improvement of efficiency of negotiation, if a learning mechanism is introduced to the multi-Agent negotiation system, so that one side on the negotiation can carry out reasoning according to the other side's belief and conduct the further learning of the other's behaving way. Therefore, learning issues in the negotiation process is gradually been paying more and more attention by researchers.This paper analyses the multi-Agent negotiation model, based on a deep research on the multi-Agent theory and RBF neutral network technology. Aiming at the lack of learning-ability about the information in the negotiation history, while, nowadays, learning algorithms are usually applied in the negotiation process, an RBF network learning is introduced to update the belief about the environment and other individuals before using Agent for negotiation, which can make the Agent conduct forecast and learning for the key result set of the negotiation in the mutual negotiation, and in that way, the reasoning strategy can be made and the efficiency of mutual negotiation and the effectiveness of individual in negotiation will be enhanced.Through the analyses and research about the multi-Agent negotiation model, RBF neutral network technology is applied to the multi-Agent negotiation and optimizes the negotiation model, which makes Agent be able to use the forecast result, according to its needs, to shorten the negotiation process, and, at the same time, acquire reasonable yields to make it easier to control the negotiation process. Based on that, BNM (Bilateral-Multi-Issue Negotiation Model, BNM)provided with learning mechanism is introduced, and defined, analyzed and designed from the aspect of negotiation model, negotiation protocol, negotiation algorithm, utility evaluation, and negotiation flow process. Finally, according to the multi-Agent negotiation model based on RBF neutral network, a simple demonstration system of electronic commerce negotiation model is finished and is proved for its feasibility through example validation.