Research On Trust Model And Related Algorithm In C2C Reputation Mechanism

Reputation mechanism is the foundation of transactions and also an effective way tohelp users establish trust relationship between each others. The core of the reputationmechanism is the trust model, at present, most C2C e-commerce sites generally useSimple Summation trust model or Average trust model but both of them cannoteffectively motivate users keep honest trading. According to this issue, the paper makessome corresponding improvement work.First of all, based on the analysis of existing C2C reputation mechanisms, the paperpropose an improved C2C e-commerce trust model. Simple Summation trust model onlytakes the rating of the transaction to compute reputation of user. Unlike the SimpleSummation trust model, the new trust model takes into some new character parameterssuch as the transaction price and transaction time. According to the two new factors, thepaper selects historical transaction information of two users within a certain period oftime in Taobao to carry out a comparative analysis between Simple Summation trustmodel and the proposed model.Secondly, learning from the trust models in P2P networks and the establishmentprocess of the trust relationship between peoples in the field of the sociology andpsychology, the paper proposes a new C2C e-commerce trust model based on therelationship of users. The proposed trust model incorporates the relationship between thetransaction requestor and the referral to evaluate the target users indirect trust value. Thenew trust model conforms to communication logic of people in the daily life and at thesame time can effectively strengthen the robustness of reputation mechanisms incombating malicious attacks such as unfair recommendations, traitors, collusivebehaviors and so on.Thirdly, most existing trust models cannot effectively prevent the fraudrecommendation behaviors. According to the issue, the paper proposes a new C2Ce-commerce trust model based on unacquainted recommenders. The new trust modeltakes Rating deviation, frequencies and risk into consideration to evaluate the credibility of unacquainted recommenders. By this way, the new trust model can more truly reflectthe ability of users to recommend.