FRAUD DETECTION AND PREVENTION IN FINANCIAL INSTITUTIONS

Financial fraud, characterized by deceptive practices aimed at securing financial gains, has emerged as a prevalent threat within companies and organizations. Traditional methods like manual verification and inspection are inaccurate, expensive, and time-consuming in identifying fraudulent activities. With the rise of artificial intelligence, machine learning-based approaches offer a promising solution to detect fraudulent transactions through the analysis of vast financial datasets. This paper presents a systematic literature review (SLR) that systematically surveys and consolidates existing literature on machine learning (ML)-based fraud detection. Employing the Kitchenham approach, which employs defined protocols to extract and synthesize relevant articles, this review reports the findings from various studies gathered through specified search strategies from electronic databases. After applying inclusion/exclusion criteria, 93 articles were selected, synthesized, and analyzed. The review outlines popular ML techniques utilized for fraud detection, prevalent fraud types, and evaluation metrics. The surveyed articles indicate that support vector machine (SVM) and artificial neural network (ANN) are commonly employed ML algorithms for fraud detection, with credit card fraud being the most addressed fraud type using ML techniques. Additionally, the paper discusses key issues, gaps, and limitations in financial fraud detection and suggests potential areas for future research. Furthermore, the review identifies challenges like class imbalance in datasets and the dynamic nature of fraud patterns. While SVM and ANN are popular, exploring ensemble methods, deep learning, and anomaly detection techniques shows promise for improving detection accuracy in suitability. Future research should address these limitations and foster collaboration between academia, industry, and regulatory bodies to advance fraud detection methods and safeguard financial transactions.