A Comprehensive Review of Machine Learning Techniques for Detecting Fraud in Banking and Payment Services

In the era of the digital world, Fraud detection in financial services has become a significant area of research because of the rising complexity of financial transactions and the advanced nature of fraudulent activities due to the enormous number of online transactions through various modes. Machine learning approaches have gained popularity in tackling these issues because of their proficiency in analyzing extensive data and detecting minor patterns suggestive of fraudulent behavior instantly through real-time analytics. This paper provides a thorough review of the use of machine learning for fraud detection in Banking and Payment services. In this paper, we will discuss the area of fraud focusing on Banking and Payment Services. We also proposed the framework of the reviewed paper that comprises 3 stages, in the first stage we formulate the research question and in the second stage we define the methodology of the research that describes the Data source, inclusion, and exclusion criteria of the present study. In the third stage, we provide an in-depth analysis of existing literature related to fraud Detection in the Financial domain. The existing literature survey indicates that supervised and unsupervised learning algorithms are the most commonly used methods for detecting fraudulent transactions. Each algorithm was evaluated in reviewed papers in terms of suitability for different types of fraud scenarios and through the various accuracy measure metrics like precision, recall, F-Score, etc. Our findings reveal that Naïve Bayes, KNN, Deep Learning, Support Vector Machine, Decision Tree, ANN,XGBoost AdaBoost models are used to detect fraud in Payment System Ensemble approaches that combine multiple algorithms are also highlighted as effective strategies for enhancing detection performance. Finally, the review concludes with suggested future research directions and emerging trends in fraud detection, such as the rapid increase in wallet payments. We propose shifting focus toward making wallet payment systems more secure by developing robust and efficient algorithms.