In view of the problems of commercial banks’ risk sources interwoven and accelerated transmission, and the recognition lag and decentralized control of traditional risk control system under the background of financial digital transformation, this paper constructs a comprehensive risk management reconstruction framework covering multi-source data governance, intelligent early warning, hierarchical control and feedback update. In this paper, credit information, trading behavior, liquidity indicators, operation logs and external public opinion are integrated into the unified calculation process, and feature engineering, machine learning recognition and rule engine linkage are combined to form a closed-loop mechanism of “identification-early-warning-control-feedback”. The experimental results show that the Accuracy of the model on the test set is 0.914, the F1 value is 0.881, the AUC is 0.942, and the Brier Score is reduced to 0.108. The accuracy of risk classification reaches 0.918, and the recall of high risk is 0.872. After dynamic control, 38.4% of high-risk samples fell back to medium risk, 12.7% fell back to low risk, the average response time was shortened to 2.9 hours, and the loss pressure drop was 23.6%. The results show that financial digital transformation can effectively improve the risk identification accuracy, control response efficiency and closed-loop governance ability of commercial banks.
