Skip to Main Content
Fraud is increasing with the extensive use of internet and the increase of online transactions. More advanced solutions are desired to protect financial service companies and credit card holders from constantly evolving online fraud attacks. The main objective of this paper is to construct an efficient fraud detection system which is adaptive to the behavior changes by combining classification and clustering techniques. This is a two stage fraud detection system which compares the incoming transaction against the transaction history to identify the anomaly using BOAT algorithm in the first stage. In second stage to reduce the false alarm rate suspected anomalies are checked with the fraud history database and make sure that the detected anomalies are due to fraudulent transaction or any short term change in spending profile. In this work BOAT supports incremental update of transactional database and it handles maximum fraud coverage with high speed and less cost. Proposed model is evaluated on both synthetically generated and real life data and shows very good accuracy in detecting fraud transaction.
Date of Conference: 28-29 Dec. 2010