Research On Multi-Agent Automated Negotiation System In Electronic Commerce Based On Trust Mechanism

Negotiation plays a decisive role in the business activities. With the rapid development of e-commerce, business behavior on the network has become more and more common. The traditional negotiations between people are unable to meet the network environment. In this context, the research of negotiation system for e-commerce has become mature gradually. But there exists a question that we didn’t have a corresponding trust evaluation mechanism for this one-multi negotiation context, this incurs the low efficiency of the negotiation. For the above reason, this thesis introduces the trust between negotiators into negotiation process, and then builds a negotiation system based on trust mechanism to solve the defect of pure negotiation.First, this thesis summaries and analyzes the existing researches on multi-agent negotiation system and points out its deficiencies. Then it presents the trust mechanism in the e-commerce environment, and introduces the trust mechanism into the negotiation system.Second, this issues proposes solutions for the problems exist in the trust research of multi-agent system, such as too little sources of trust information and not precise enough of the trust information processing. In terms of the information achievement, previous models are focus on the direct experience, witness information which the evaluator collect experience of its acquaintance that interacted with the target agent, third-party references provided by the target agents. Our model add a new source that provider will show consumer the interactive records of its competitors around. In terms of the information processing, previous models define the weight of acquaintance information according to the distance to the provider. In the recommendation system, the weight of acquaintance is related with the similarity between consumer and acquaintance. And then, in our model, we presented that the weight of acquaintance information should be set related to correlation between the consumer and acquaintance. Based on the considerations above, we establishes a new multi-agent model for trust information achievement and processing, which improves its accuracy and makes it more adaptive to open dynamic multi-agent environment.Third, this thesis combines the time dependent tactics and behavior dependent tactics of the negotiation. The consumers will not give the proposal just based on the time consumption, but also imitating the providers’ behavior according to their proposals. And then, we propose the negotiation protocol based on trust mechanism, describes the architecture and communication mode between consumer and provider. At last, a negotiation system based on trust mechanism has been designed.Fourth, in the platform of swarm, we compare the negotiation results based on trust mechanism and no-trust mechanism under the static environment and dynamic environment separately. Our results prove that negotiation based on trust mechanism can greatly improve the negotiation results.Our thesis improves the original trust mechanism and combines it with auto negotiation system, which makes the auto negotiation system more effective and adaptive to the e-commerce environment.