The Evolution of Financial Fraud Detection Methods: A Systematic Review of Integration of Theory, Data Analytics, and Artificial Intelligence

Abstract

Financial fraud is a persistent global threat that undermines the reliability of financial reporting, corporate governance, and economic stability. In Indonesia, recent high-profile cases such as the LPEI corruption scandal illustrate the limitations of existing fraud detection systems in identifying complex and concealed fraudulent behavior. The growing sophistication of fraud patterns, coupled with increased data volume and the digitization of financial systems, presents a significant challenge to traditional, manual-based detection methods. This highlights a critical gap in both theory and practice regarding how fraud is detected, interpreted, and prevented. This study aims to analyze and describe the evolution of financial fraud detection methods over the past decade and examine the role of Machine Learning (ML) and Explainable Artificial Intelligence (XAI) in enhancing accuracy and trust in financial fraud detection systems. A systematic literature review was conducted using the PICO framework, focusing on peer-reviewed articles published between 2019 and 2024 sourced from the Emerald Insight database. The results show a clear transition from traditional fraud detection approaches such as document analysis, field investigations, and interviews toward automated, data-driven techniques. The integration of ML algorithms, including Support Vector Machines, Random Forests, and unsupervised clustering, has improved fraud identification accuracy. Additionally, the use of XAI enhances model interpretability and stakeholder confidence by addressing the black-box nature of AI models. These technologies not only streamline detection processes but also reduce false positives and improve decision-making transparency. This research contributes to the literature by mapping the convergence of behavioral fraud theories and data science approaches. It also offers practical insights for organizations and auditors in developing adaptive, technology-integrated fraud detection frameworks that are both accurate and explainable.