Big Data-driven Financial Risk Prevention and Control Model for Commercial Banks

Abstract

With the rapid digitalisation of financial markets, many new kinds of complex and interconnected risks for commercial banks have emerged that cannot be controlled by the old ways of risk management. Big data technology can help to better solve the problem of risk prevention and control in finance by collecting a large amount of data, analysing it, and issuing timely early warnings. A complete big data-driven financial risk prevention and control model for commercial banks has been put forward in this paper, and it is composed of many analysis modules to address problems such as credit risk, market risk, liquidity risk, operating risk, fraud detection, etc. A group of gradient-boosted decision trees is used for credit risk, LSTM is employed to predict market and liquidity risk, GNN is applied to detect fraud and operational risk, and all of them are combined to obtain a composite risk score and a five-tier early warning classification. Based on the empirical test, the integrated model has an AUC-ROC of 0.963 and an F1-score of 0.924, and it has performed better than all traditional statistics and single-module machine learning methods for the risk categories in question. Based on SHAP feature importance analysis, the main causes of credit risk are payment delinquency history and debt-to-income ratio. There are also problems with the application of data governance, model explainability and regulations in this study, and some directions for future research have been put forward, such as federated learning and causal inference.