Introduction The intersection of big data analytics and financial risk management has spurred significant methodological innovation and organizational change. Despite growing research activity, the literature remains fragmented, with notable gaps in comparative effectiveness, cross-sectoral applicability, and the use of non-traditional data sources. Methods Following the PRISMA 2020 protocol, a systematic review was conducted on 21 peer-reviewed studies published between 2016 and June 2025. The review evaluated the methodological diversity and effectiveness of machine learning and hybrid approaches in financial risk management. Results The analysis mapped the relative strengths and limitations of neural networks, ensemble learning, fuzzy logic, and hybrid optimization across credit, fraud, systemic, and operational risk. Advanced machine learning techniques consistently demonstrated strong predictive accuracy, yet real-world deployment remained geographically concentrated, primarily in Chinese and European banking and fintech sectors. Applications involving alternative and unstructured data, such as IoT signals and behavioral analytics, were largely experimental and faced both technical and governance challenges. Discussion/conclusion The findings underscore the scarcity of systematic benchmarking across risk types and organizational contexts, as well as the limited attention to explainability in current implementations. This review identifies an urgent need for comparative, cross-jurisdictional studies, stronger field validation, and open science practices to bridge the gap between technical advances and their operational impact in big data–enabled financial risk management.
Theodorakopoulos et al. (2025) studied this question.