MACHINE LEARNING TECHNIQUES IN DETECTING FRAUD: A STATE OF THE ART REVIEW IN THE SCOPUS DATABASE

Abstract

This study aims to understand the development and impact of machine learning in fraud detection. The study was conducted by reviewing 106 scientific articles indexed in Scopus and published between 2009 and 2025, specifically in the fields of Business, Management, Accounting, Economics, and Finance. The results indicate that interest in the use of machine learning in fraud detection has increased, especially after 2020. Compared with traditional methods, machine learning techniques can provide more accurate and faster results. However, challenges remain, such as the difficulty of understanding how the models work, imbalanced data, problems in developing widely usable systems, and their direct application. Furthermore, research in this field is still scattered across many disciplines and lacks clear assessment standards. Therefore, interdisciplinary collaboration, the development of explainable AI systems, and access to real-world data are needed. In conclusion, machine learning has great potential to improve fraud detection, but it still requires a more targeted, ethical, and practical approach to achieve widespread and relevant use across a wide range of situations. These findings provide an important foundation for future research and development of fraud detection technologies.