Analysis of Banking Fraud Detection Methods through Machine Learning Strategies in the Era of Digital Transactions

Abstract

Nowadays, electronic banking activities are becoming increasingly popular and are projected to become much more prevalent as digital financial transaction systems advance. One unintended consequence of this development is that fraudulent transactions have become a significant issue in online banking. As technology advances, so do criminals’ means of perpetrating fraud. There are also growing technologies that allow fraudsters to replicate the transaction behavior of actual consumers, and their tactics are constantly evolving, making it harder to identify fraud. Dealing with that, banks and financial institutions have deployed several models based on various approaches, including ML (Machine Learning), DL (Deep Learning), and RL (Reinforcement Learning), to identify fraudulent activity, considering the advent of big data. In this study, first, we present banking fraud cases. Afterward, we provide an overview of banking fraud detection techniques such as supervised, unsupervised, optimization, DL, RL, and hybrid approaches. Ultimately, we identify the strengths and limitations of the existing study and highlight some recommendations and future scope.