Study On Reputation Model For Online Auctions

With the development of internet, e-business and auction websites grow promptly. Online auction, one of the components of e-business, plays an important role in promoting optimized resource distribution and accelerating the process of economic globalization. However, serious integrity problems occur when the world is exploring and developing online auctions as an emerging negotium, which becomes a bottleneck for online auction development, having an influence on healthy, orderly and rapid development of social economic activities directly or indirectly. Reputation model weighs heavily in establishing and maintaining online trust relation and guaranteeing the order in online market.We find that these problems that situational characteristics of online trade are not taken fully into account, uses'credit standing is not reflected completely and honestly and convenience for using does not exist, etc. occur for existing reputation models after the comparison and analysis for those mainstream from ways of reputation rating model, reputation model parameters, frauds prevention and so on. And we put forward basic conditions that a reputation model having good performance should satisfy and analyze every parameter of the model. Then we improve on Additive Reputation Model which has been used widely at present, introducing these parameters reputation feedback score, transaction value and evaluation time to the model and confirm weighting coefficients of parameters. And Narrow Logistic Growth Curvilinear Model used widely in economics forecasting is applied when we calculate the weight of transaction value.We prove that the reputation model is efficient by empirical study based on real transaction data of Taobao. It is typical to take Taobao who is dominant in personal trade market of our country for example. At the same time, it is good sample for studying microcosmic economic problems such as online auction and reputation because online transaction has observable behaviors and measurable features. To get the sample, we develop a program, Automatic Network Information Acquisition System, with MS VS.Net 2005 and MS SQL Server 2005 storing non-structured data from plenty of web pages to structured database. At last, by calculating sample users'reputation scores with the reputation model presented by us, a conclusion is drawn that the model is better than Additive Reputation Model. To some extent, honest and fraudulent users are distinguished with the former model, which can prevent ill speculation effectively, make it equitable and reasonable to calculate users'reputation scores and then promote the development of online auctions.