A Systematic Review Of Credit Card Fraud Detection And Prevention Techniques

Abstract

The data science community has shown significant interest in credit card fraud (CCF) due to its growing prevalence and the significant financial losses it causes. However, existing systematic reviews have not thoroughly explored the various techniques used in CCF beyond mere comparisons.
The objective of this research is threefold. First and foremost, it aims to provide a clear definition and classification of CCF. Furthermore, it aims to provide a comprehensive overview of the standard techniques currently employed to prevent CCF, ensuring strict adherence to industry and international payment network standards. The study also seeks to explore alternative tools proposed in existing literature. Lastly, the research aims to conduct a systematic analysis of advanced techniques using different aspects of machine learning and deep learning models. This analysis includes considering the type of model, estimation metrics, model comparison, as well as the challenges encountered during the process.
A comprehensive search was conducted across multiple electronic databases to review studies published between 1990 and 2021. The review concentrated on two primary categories: studies focused on the classification of credit card fraud and those that examined the key techniques used for its prevention and detection.
Deep learning (DL) models show great potential in the detection and prevention of CCF. However, their use in industry remains limited, highlighting the need for further efforts and incentives to support their implementation. Based on our review, we offer targeted recommendations for researchers and practical guidelines for industry professionals.