Online Credit Card Fraud Detection and Anomaly User Blocking

Abstract

The use of credit cards has expanded considerably as a result of rapid advancements in electronic commerce technologies. As credit cards become the most popular method of payment for both online and offline purchases, incidences of credit card fraud are on the rise. Transactions with Credit Cards continue to expand in number, capturing a larger share of the US payment system and resulting in a higher rate of stolen account numbers and bank losses. Improved fraud detection has thus become critical to the payment system's long-term stability in the United States. For several years, banks have deployed early fraud warning systems. Large-scale data mining tools have the potential to improve commercial practice. Scalable strategies for analysing enormous volumes of transaction data and efficiently computing fraud detectors in a timely way, especially for e-commerce, is a critical issue. Aside from scalability and efficiency, the fraud-detection task has technical issues such as skewed training data distributions and non-uniform cost per error, both of which have received little attention in the now-established ledge-discovery and data mining communities.