Literature Review of Different Machine Learning Algorithms for Credit Card Fraud Detection

Abstract

Every year fraud cost generated in the economy is more than $4 trillion internationally. This is unsurprising, as the return on investment for fraud can be massive. Cybercrime specialists estimate that an investment of 1 million dollars into fraud or attack can net up to $100 million. Financial institutions such as commercial and investment banking operations are increasingly being targeted. And we know that the only way to fight fraud effectively is through the use of advanced technology. The answer lies in relying on advanced analytics and enterprise-wide data storage capabilities that support the use of artificial intelligence (AI) and machine learning (ML) approaches to stay one step ahead of criminals. AI is best suited to defend against today’s fast-changing and complex bank fraud, where new threats are under development every day. Approaches relying on fragmented and siloed data, rules-based approaches or traditional point-solutions are no longer acceptable. These approaches are not only ineffective, but they are extremely costly to banks and financial services firms because they force legal and compliance teams to spend a lot of time trying to gain access to the data they need. By relying on advanced analytics and AI and ML capabilities, fraud and compliance units can spend their time working on more-complex fraud issues. Manual investigation can be reduced through the use of complex algorithms powered by ML, often in conjunction with rules, a combination that offers significant advantages over purely based -rules fraud detection. In this paper, we have included different machine learning algorithms used to detect credit card frauds and also provide a comparative study between different algorithms.