Collusive Fraud Classification in Network of Online Auction Using Similarity Measure in Collective Classification

Nowadays, data classification is extremely important used with the purpose of identifying the features that indicate the group of the classification of each item. Classification of the user auctions is one of the usages of classification. In previous years, electronic auctions have become more important, so detecting fraudulent activities has attracted attention of many researchers. One type of fraud is the collusion of  fraudulent users at the auction, which is a very dangerous type of fraud and if occurred, may lead to       irreparable financial losses. In this paper, we propose a method that first extracts the effective features for finding normal people in the auction and then classifies the users by collective classification method. We define an edge potential function to use in collective classification, in which it uses the distance L1-norm as the similarity measure between the two adjacent nodes. The results show that the defined edge potential function is suitable for improving the classification rate of collaborative fraudulent users.