A k-means clustering based algorithm for shill bidding recognition in online auction

A new method based on k-means clustering is proposed for the shill bidding detection in online auction. Through analyzing the behavioral characteristics of buyers, the proposed method extracts and quantifies four characteristics for every buyer, that is to say each buyer will be represented by a vector of four elements. Then all buyers are divided into two categories, i.e., shill bidding buyers and general buyers by the proposed k-means clustering based algorithm. An example that collects actual data of an online auction from one online store and then analyzes the data with SPSS is given to show that two types of buyers differ significantly on the four characteristics. The results illustrate that this new method is effective and suitable to be generally used.