The role of big data analytics in financial fraud detection: A systematic literature review

This literature review examines existing research on the application and development of Big Data Analytics (BDA) in financial fraud detection and its broader implications for market transparency. The purpose is to investigate the main users of BDA, its approaches and tasks, as well as future research topics and trends related to fraud detection. The main users of BDA are auditors (46%), analysts (27%), investors (19%), and regulators (8%). The most widely used approaches, namely predictive analytics (54%), hybrid frameworks (34%), and visual analytics (12%), indicate a tendency to rely on historical data in detecting financial fraud. Meta-analysis identifies four main BDA applications in financial fraud detection: supervised classification, anomaly detection, linguistic feature extraction, and unsupervised textual pattern discovery. Trends and future research directions related to the use of structured data should align with data characteristics. When dealing with imbalanced datasets, selecting methods tailored to the data characteristics is essential. Unstructured data enhances accuracy in financial fraud detection when combined with financial data. This literature review has methodological and theoretical implications. The use of structured and unstructured data requires attention to appropriate BDA approaches and tasks. Behavioral theories, such as the fraud triangle and fraud pentagon, can be integrated with BDA to detect financial fraud, with predictive performance strongly shaped by the quality of the proxies employed.