Fraud Identification On B2B E-commerce Platform

B2B is a market where opportunity and risk coexist and will bring business opportunities as well as greater credit risk to enterprise. Root lies in the information asymmetry of both parties. Because of a B2B transaction amount is large, many enterprise users are not willing to take a risk to trade with strangers on an e-commerce platform. The short of network trust becomes the main reason for the development of B2B e-commerce. Internet fraud is a important factor that raised the lack of trust. The small conditionality of information published on the net, as well as the virtually and imperceptibility of the network make the consumers to judgment information incorrectly, even hard to find out the enterprises after detect the error information. Therefore, some companies acting reckless, make all kinds of false information on the Internet, or used to make a variety of news to attract consumers, to expand their business impact, seeking economic benefits. The proliferation of false information, to a certain extent affected the trust of consumers shopping over the Internet. E-commerce makes social credit issues become more prominence, deception and fraud, happening from time to time, restricting the development of e-commerce, become a problem needed to resolve. Thus, effectively identify Internet fraud has become particularly important.In this dissertation, we analyzed the fraud in e-commerce research and it’s characteristics. Then described the method of data mining technology and its application in the field of fraud monitoring. Analyzed data mining classification method, and through the examples analysis to choose the good classifier-Random Forest for empirical research. Building the users classification model with user data of B2B e-commerce platform called made-in-china, in order to effectively recognition the fraud users. Innovation of this dissertation lies in using the R statistical analysis software combined with random forest building user classification model and verify the validity of these models, effectively identify fraud users and solve real-world problems.