AI for financial fraud detection: A hybrid deep learning framework

Akavaram, Sravanthi (2025) AI for financial fraud detection: A hybrid deep learning framework. World Journal of Advanced Engineering Technology and Sciences, 15 (2). pp. 2626-2633. ISSN 2582-8266

This article presents a hybrid AI-driven architecture for real-time detection of financial fraud across high-volume transactional networks. Leveraging graph-based anomaly detection, temporal deep learning models, and adaptive learning, the proposed framework identifies complex fraud patterns including synthetic identity fraud, account takeover, and multi-account collusion networks. Traditional rule-based systems struggle with high false positive rates and slow adaptation to novel fraud patterns, whereas this hybrid model combines Graph Neural Networks, Temporal LSTM Networks, Autoencoders, and Adaptive Boosting to create a comprehensive detection system. Key innovations include FraudNet for identifying relational anomalies, Time-Aware Autoencoders for temporal pattern recognition, Real-Time Reinforcement Learning for continuous adaptation, and Multi-view Fusion for integrated analysis. The framework has been validated through real-world implementations across multiple financial institutions, demonstrating substantial improvements in detection accuracy, reduction in false positives, and efficiency in operational processes while maintaining millisecond-level latency for real-time transaction processing.