Credit Card Fraud Detection Methods: A Review

Departement of Computer Science & Engineering., Madhyanchal Professional University Ratibad, Bhopal, M.P., INDIA

^* Corresponding author: chetanhpatil@gmail.com

Abstract

In today’s context, the term “fraud” has become closely intertwined with credit card-related deceit. Recent years have witnessed a notable surge in both credit card utilization and fraudulent activities. Detecting and thwarting fraud necessitates a meticulous analysis of customers’ spending patterns. The ubiquity of credit card use for both online and in-store transactions has un-fortunately led to a parallel rise in recognition valentine scam occurrences. While the primary area of deception discovery is the documentation of sham incidents, the urgency of promptly flagging such events cannot be overstated. The con-temporary landscape heavily favors credit card usage, a trend that inadvert-ently contributes to the annual expansion of fraudulent gains. This unlawful practice exerts a pernicious influence on the global economy at large, exac-erbating its impact year after year. Numerous cutting-edge methods, including as data mining, machine learning, fuzzy logic, genetic programming, sequence alignment, artificial intelligence, and fuzzy logic, have become indispensable in the fight against this threat when it comes to identifying credit card fraud. This study delves into the intricate integration of data mining methodologies, showcas-ing their robust potential to provide comprehensive coverage against fraudu-lent activities while maintaining a controlled balance between false alarms and detection accuracy. Within the financial sector, the challenge of credit card fraud detection remains both persistent and pressing. This paper intro-duces an innovative paradigm aimed at fortifying credit card fraud detection. This is achieved by synergistically harnessing the capabilities of the Artificial Underground Over-sampling Practice (SMOTE), the potency of Adaptive Increasing (ADABoost), and the privacy-enhancing attributes of Federated Learning. The incorporation of federated learning serves a dual purpose: not only does it address prevailing data privacy concerns, but it also significantly augments the precision of fraud detection across a diverse array of geograph-ically distributed data sources