RISK ASSESSMENT IN CLOUD-BASED BANKING USING TRANSFORMER-CNN HYBRID MODELS AND BAYESIAN NETWORKS

Computer Engineering

Today, the complexity of the finance and the eccentricities of the financial system at a personal level requires an extremely strong models for fraud detection and risk assessment. To this end, this research work propounds a Federated Learning Financial Risk Prediction model realized with Transformer-based deep neural networks and Bayesian Networks to increase the accuracy of fraud detection at the same time whilst preserving privacy. Federated Learning has the decentralized and secured data processing while protecting highly sensitive financial variables. Promising enhanced performance through a Transformer-based architecture, since these networks will extract very short and long dependencies present in transaction data. Its application is also covered by probabilistic estimates of risks by minimizing false positives and false negatives. The outcomes produced from the experiments provide excellent results in accuracy, precision, sensitivity, and specificity that prove the model's efficacy in predicting financial risk. It stretches beyond any limitation for applicability over huge datasets for the finance domain. The future work intends to take up the improvement of approaches in Federated Learning, design of quantum-safe cryptographic methods that are quantum-resistant, and risk management of finances using blockchain technology to ensure security and transparency in financial systems.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.