A Multiple-Phased Model for Early Fraud Detection in Online Auctions Using a Neural Networks Ensemble Classifier

A growing number of people is performing electronic transactions through online auctions. However, auction frauds increasingly threat electronic auction websites. Many online auction sites provide reputation systems to prevent fraud through rating the traders and giving buyers the chance to evaluate a seller's credit. However, Fraudsters can easily manipulate these feedback score solutions by creating fake overrated reputations. Thus, designing an effective frauddetection model can help traders of inline auction to identify cases of fraud. However, the most important issue of early fraud detection, which is identifying a fraudster before he commits a scam, has been examined in very few of the previous research works. This paper proposes a new phased detectionmodel to discover potential fraudulent activities in online auctions as early as possible. In order to improve performance in detection, we use a neural network ensemble which learns the behavior models of the fraudsters in different phases and combines the predictions of individual models. Theexperimental results on real transaction data collected from eBay for training and testing show that the proposed model offers a good performance for fraud detection and recall and precision rate of fraud detection is high.

کلیدواژه ها:

online auctions ، fraud detection ، ensemble classifier ، phased modeling