On Optimizing Model Of Trust And Negotiation In Agent Based Online Collaboration

With the rapid development of Internet technology and wide application of computers, E-commerce based on Internet have developed correspondingly, which also brings new challenges and opportunities to the traditional business,Multi-agent systems play an important role in E-commerce.However, there exist two shortages in current agent technology that need to be improved for satifying the reguirment of E-commerce system. The first one is that there may exist cheating phenomenon in the online collaboration, which leads to lower cooperation success rate and threatens the reputation of E-commerce. Thus the trust model needs to be improved. The second one is that due to the dynamic environment, the current negotiation techniques are not flexible enough to meet the needs of internet-based e-commerce system. This paper presents an improving scheme to these issues from the perpective model of trust and intelligence.Firstly,from the aspect of Agent system on trust, this paper proposes a trust optimal model and a selection algorithm. The model which called DRI trust model, it built based on the third-party management method in E-commerce. And the algorithm is used for cooperative object in the frame of DRI trust model. Trust experimental results show that the average cooperation success rate of DRI trust model is higher than one of tranditional trust model. The main reason of the results is the cheating phenomenon in E-commerce has been suppressed in a certain degree.Secondly, from the aspect of Agent system on negotiation, reinforcement learning theory is introduced in this paper. On the basis of Watkins'Q-Learning theory, a Q-Learning negotiation strategy is proposed and implemented. Experiments prove that the Q-Learning negotiation strategy has optimized the negotiation process and enhances the negotiation efficiency of the Agent online cooperation.Finally,the paper takes a example for testing online cooperation between traveler'Agent and travel agency'Agent respectively. It proves DRI trust model and Q-Learning negotiation strategy in E- commerce is rational and effective.