Optimizing Financial Security with Cloud AI: Implementing Deep Q-Network and Transfer Learning for Risk Management and Fraud Detection

Fraud detection and risk management remain to be significant issues in financial security due to constantly changing threats. This research introduces an innovative method that integrates transfer learning with Deep Q-Networks (DQN) for effective financial anomaly identification. The suggested approach boosts detection capabilities by utilising pre-trained models for feature extraction and reinforcement learning for decision-making. The system surpasses conventional methods, attaining 98.7% accuracy, 97.8% F1-score, and 99.1% AUC-ROC on benchmark datasets. In comparison to baseline models like Random Forest and XGBoost, the suggested method shows enhanced accuracy and computing efficiency. These findings confirm the framework's capability for real-time fraud prevention and dynamic risk assessment.