A Multiple-Phased Modeling Method to Identify Potential Fraudsters in Online Auctions

Loopholes in online auction sites enabled fraudsters to easily hide themselves. To reduce the odds of being defrauded, online auction traders usually use reputation systems for estimating a trading partner´s credit. However, reported dollar losses of online auction fraud have hit recorded height for years that implies existing reputation systems may not prevent fraud effectively as expected. To reduce the risk of being defrauded, an ideal fraud detection mechanism should be not only to identify current fraudster but also potential ones. Therefore, this paper proposes a multiple-phased modeling method integrating with decision trees for enhancing the capability of fraud detection. To demonstrate the effectiveness of the proposed method, real transaction data were collected from Yahoo! Taiwan for training and testing. The experimental results show that the recall rate of identifying a potential fraudster before transitioning into his criminal phase was up to 86%.