Research Of Multi-Agent Negotiation Mechanism Based On Bayesian Learning

With the integrating of Agent technology and E-commerce, e-commerce based on Agent technology has become the hot spots of Academic research. Currently e-commerce did little towards Business automation technology, particularly the domain of automated negotiation. The characters of Autonomy, sociality, reactivity and self-learning of Agent adapted to the flexibility of e-commerce. Agents on behalf of the negotiators reach a Win-win goal by self-learning and changing proposing.How to improve the negotiation ability is the key problem in the study of Multi-Agent System. Theoretical analysis shows that it can complete negotiation goal by improving the learning mechanism in Multi-Agent Negotiation System and making every Agent coordinate their own behaviors through learning. Therefore, the application of machine learning theory to automated negotiation system has become the latest e-commerce research.This paper studies that how to improve the efficiencies of negotiation in the bilateral problems. At first, analyzing the limitation of traditional single negotiation strategies of Time-based, resource-based and behavior-based and so on, we put forward a proposal based on the combination of the negotiation strategies, adopting the combination algorithm of weighted sum method to combine the different single negotiation strategies. Then we expatiated on Bayesian learning in the multi-Agent negotiation, and designed an E-commerce negotiation model based on Bayesian learning, decrypting the negotiation process in detail. At last, we designed model framework of the E-commerce negotiation system, and preliminary implement an E-commerce negotiation system based on learning.