Research On The Optimization Of Trust Feedback System In Online Auction

Recently,with the rapid development of information technology, online auctions has become more tremendous. As a new style of trade, the online auction not only change the traditional trading pattern of human society, but also promote the auction from the aristocratic civilian transformation. However, with the enjoyment of freedom and convenience of online auction, there are more and more fraud simultaneously. Because of the asymmetry between buyers and sells in information,false trading, counterfeiting, phishing, bidding up,more snd more illegal activities happen in online acution. The lack of online trust has become an important factor hindering the rapid and healthy development of online auction. Therefore, it is particularly important to study the credibility- the only credit reference index . The design of a scientific and rational credibility of feedback system setting up for the online auction transaction protection mechanism of trust perfectly, and promoting the healthy development of online auction.The thesis mainly adopts the theory and the empirical research on online auction to model credit feedback system and analyse its validity. Firstly, the characteristics of online auction and the influencing factors of hindering its development were investigated. Through analyzing the shortcomings and the insufficiency of credit feedback system , the thesis take trade value, the factors affecting the credibility rating etc. into the model, and subdivide the evaluation. The Sporas model is improved . The better models were built by two ways to calculate users'credibility weight . The two ways are user's reputation score and the collaborative filtering .The users are divided into five different types :good faith , no conspire and cheat stability , no conspire and cheat unstability, conspire and cheat stability ,conspire and cheat unstability .. Two existing evaluation models and models built in this article as a four-action programs. Trust calculation error as a decision-making matrix, of the four-level model to determine the level of the use of ELECTRE-Ⅰevaluation methods to evaluate the model in order to test the credibility of the feedback to optimize the effectiveness of the evaluation model.Through the study, the thesis got the following results and findings:1, Establish a new evaluation model of the credibility .First of all, the credibility of the buyers'feedback should be subdivided. The buyers'evaluation should be multi-faceted, and the weight of this factors are different . These factors should be evaluated and given appropriate weight coefficient. This will enable the buyers'evaluation more comprehensive and more accurate. Secondly, the credibility of the user's score should not be simply additive . It is influenced by the score of users, the transaction price, the transaction time. In the thesis , the factors were took into the model as the weight factor of the credibility. Finally, taking the two points into account, establishes a new model of the credibility of the feedback.2, Verified the validity and usefulness of the model.Then, through the introduction of trust calculation error (TCE) to the quantitative calculation proved the model's practicality and effectiveness. The users are divided into five different types ,which can be as five status, four types of model as four different actions . Then using the ELECTRE-Ⅰevaluation method simulates . We got the conclusions as follows:Finally, the two models built in the thesis were proved better than the existing models. Among them, when the system exist the two types of users ,no conspire and cheat stability and no conspire and cheat unstability , the model based on user's reputation score can be better than any other model. When the system exist the two types of users , conspire and cheat stability and conspire and cheat unstability , the model based on the collaborative filtering can be play its role better than any others.3, Set up a new reasonable user credibility.Set up a dynamic new user credibility, to prevent a malicious user to abandon the credibility of information on the conduct of re-registration effectively . To some extent, both parties are motivated to trade with each other faithfulness.