Deep Learning-Driven Financial Fraud Detection: An Enterprise Risk Analytics Framework for Real-Time Anomaly Detection and Regulatory Compliance
The American Journal of Engineering and Technology · 04.06.2026 · 08 (06) · 38-63
DOI: 10.37547/tajet/volume08issue06-02 ↗Kurz gefasst
Fachartikel in der Zeitschrift The American Journal of Engineering and Technology, online erschienen am 04.06.2026. Verfasst von Shuvo Ranjan Das und 3 weiteren Autor:innen.
Abstract (Originalsprache)
The high growth rate of digital financial ecosystems has greatly amplified the magnitude, speed, and sophistication of fraudulent transactions that have presented major challenges to the conventional fraud detection systems. The traditional rule-based and statistical models usually have high false positive rates, slow detection, and minimal capability to adjust to changing patterns of fraud. In this study, it is proposed to develop a deep learning-based enterprise risk analytics framework that would allow detecting financial fraud in the real-time and meet the regulatory compliance requirements. The architecture combines sophisticated deep learning models, such as Long Short-Term Memory (LSTM) networks, autoencoders and graph neural networks, with business risk management applications, such as dynamic risk scoring, anomaly detection pipelines, and compliance monitoring layers. The proposed system is tested using the publicly available transactional datasets, like the European card fraud data, on the basis of the most important performance indicators, including accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (ROC-AUC). The results show that deep learning models by far exceed traditional machine learning methods by being more accurate in detection and significantly lowering false positives in highly imbalanced datasets. Additionally, explainable AI methods improve model transparency, which can be easily accepted by regulators and audited. The research will add to the body of knowledge by filling in the gap between superior artificial intelligence methods and risk governance models on an enterprise level and provide a flexible and scalable answer to the contemporary financial institutions. The offered framework both enhances the ability to detect fraud and facilitate proactive risk management and compliance in more sophisticated financial settings.
Zitieren
APA
Das, S. R., Afroz, S., Rashid, H. U., & Chowdhury, M. A. (2026). Deep Learning-Driven Financial Fraud Detection: An Enterprise Risk Analytics Framework for Real-Time Anomaly Detection and Regulatory Compliance. The American Journal of Engineering and Technology, 08(06), 38-63. https://doi.org/10.37547/tajet/volume08issue06-02
Vancouver
Das SR, Afroz S, Rashid HU, Chowdhury MA. Deep Learning-Driven Financial Fraud Detection: An Enterprise Risk Analytics Framework for Real-Time Anomaly Detection and Regulatory Compliance. The American Journal of Engineering and Technology. 2026;08(06):38-63. doi:10.37547/tajet/volume08issue06-02
BibTeX
@article{Das2026,
author = {Das, Shuvo Ranjan and Afroz, Sadia and Rashid, Hasib Ur and Chowdhury, MD Al-Amin},
title = {Deep Learning-Driven Financial Fraud Detection: An Enterprise Risk Analytics Framework for Real-Time Anomaly Detection and Regulatory Compliance},
journal = {The American Journal of Engineering and Technology},
year = {2026},
volume = {08},
number = {06},
pages = {38--63},
doi = {10.37547/tajet/volume08issue06-02}
}
EUVIMED