Chunk-Based Incremental Classification of Fraud Data

Shill Bidding (SB) is still a predominant auction fraud because it is the toughest to identify due to its resemblance to the standard bidding behavior. To reduce losses on the buyers' side, we develop an example-incremental classification model that detects fraudsters from incoming auction transactions. Thousands of auctions occur every day in a commercial site, and to process the continuous rapid data flow, we design a batch-based incremental classification algorithm that addresses the imbalanced and non-linear learning. We train the proposed algorithm incrementally with several SB training batches and concurrently assess the performance of the new learned models with unseen batches.