| Summary: | 碩士 === 淡江大學 === 資訊管理學系碩士班 === 101 === In recent years, the rapid growth of online auctions were seen by everyone. The convenience, concealment and not constraints by time and space, is very helpful to raise the trading volume. However, many fraudersters start to obtain illegal benefits when facing such a vigorous trading platform. The ways of fraud are not only diverse but also changing by time and environment, difficult to avoid. In order to provide a more secure trading environment, our research development a online auction fraud early detection methods based on the analysis of behavior. First, we focus on segmentation of transaction history of fraudersters and normal users by trading events, and then proceed cluster analysis to conclude typical trader state. Second, in order to create the temporal behavior associated with the classification model we segment the transaction history by trader''s state. Besides, we user the dataset that segment by trader''s state to produce the state label string, and generate sequential pattern base to help the users monitor and compare the suspicious accounts. According to the methods above, our research implements a simple online auction trading decision support system. So the users can observe and analyze the behavior of account before trading. Last, to verify the effectiveness of our proposed method, we use actual transaction history on auction site to proceed experiments. The results show that the proposed method actually helps improve the early detection of auction fraud and promote the safety of online auction trading.
|
|---|
