Clustering and Labeling Auction Fraud Data

Abstract

Although shill bidding is a common fraud in online auctions, it is however very tough to detect because there is no obvious evidence of it happening. There are limited studies on SB classification because training data are difficult to produce. In this study, we build a high-quality labeled shill bidding dataset based on recently scraped auctions from eBay. Labeling shill biding instances with multidimensional features is a tedious task but critical for developing efficient classification models. For this purpose, we introduce a new approach to effectively label shill bidding data with the help of the robust hierarchical clustering technique CURE. As illustrated in the experiments, our approach returns remarkable results.