Volume 17, Number 4
A Systematic Review of Applications in Fraud Detection
Authors
Hashim Jameel Shareef Jarrar, Middle East University, Jordan
Abstract
The following systematic review aims to investigate the applications of data science techniques for fraud detection (FD), especially Machine Learning (ML),Deep Learning (DL), and the combination of both techniques in different domains, including credit card fraud and cyber (online) fraud. The increasing sophistication of fraudulent activities necessitates advanced detection methods, as traditional rule-based techniques often fall short. The review involves articles from 2022 to 2024, establishing various algorithms and techniques' efficiency. Some of the research findings show that the most frequently used FD algorithms are supervised ML algorithms like logistic regression, decision trees, and random forests, which have high accuracy. Also, DL techniques especially Long Short-Term Memory (LSTM) networks and convolutional neural networks (CNNs), have been reported to provide better results, especially in real-world problems, including e-commerce and online web-based FD. Some of the new trends that are increasingly being incorporated to improve FD capabilities are the hybrid models that integrate ML and DL methods. However, there are still some limitations associated with the use of ML for FD, such as class imbalance, interpretability of the trained model, and the evolving nature of fraud tactics. The review discusses the current trends, including real-time detection and the use of AI in FD systems; the review also provides further research directions for overcoming the challenges and improving the performance of FD systems. Overall, this review contributes to the growing body of knowledge in FD and emphasizes the importance of continuous innovation in data science applications.
Keywords
Data Science; Machine Learning; Deep Learning; Fraud Detection; Cyber Fraud