Cloud-Assisted Batch Learning for Financial Risk Detection Using LSTM, Transformer, and 1D-CNN

Cloud-based banking systems' financial risk detection continues to be a pressing challenge because of the sophistication of fraudulent schemes. Old rule-based and static anomaly detection methods find it hard to keep up with developing fraud patterns. This paper introduces a Cloud-Assisted Batch Learning Framework that combines LSTM, Transformer, and 1D-CNN for advanced financial risk detection. The architecture uses Markov Decision Processes (MDP) for adaptive fraud detection, Blockchain authentication for safe transactions, and Reinforcement Learning (RL) for ongoing adaptation. Experiments on the Pay Sim dataset show that the proposed architecture has a high precision (98.73%) compared to traditional machine learning approaches. The false positive rate (FPR) declines to 0.5859%, and the false negative rate (FNR) is 0.4591%, showing enhanced detection reliability. These findings confirm the efficiency of the proposed deep learning-based fraud detection system as a viable solution for secure, scalable, and adaptive financial risk management in cloud environments.