A Study On Risk Rating Of Commercial Bank Based On Machine Learning Model Fusion

With the globalisation of the economy,not only are the lifelines of the entire world economy linked,but the financial sector worldwide is also affected by each other.For the banking industry,which is crucial in the financial sector,various risks caused by operational problems and the slowing down or even failure of the flow of funds to reach the accounts may,like the butterfly effect,trigger a financial crisis that will endanger the entire world and cause the collapse of the entire world financial order.Therefore,how to prevent the occurrence of risks in commercial banks and how to better supervise them are major challenges for China's banking industry and regulatory authorities.As the Western world has an earlier economic development and a more sophisticated financial system,there is a comprehensive rating system for commercial banks and a more effective management approach to various problems.In contrast,China's economy is currently developing at a rapid pace,but the risk management approach to commercial banks cannot meet the ever-changing business model,resulting in many problems in the supervision of commercial banks.The rating system for commercial banks at home and abroad is constantly being updated and optimised,and has become an important tool for observing the current state of commercial bank operations and preventing bank risks.Based on the rating results,customers and investors can have a more intuitive understanding of the bank's operation and assets;banks can prescribe the right remedy to reduce bank risks;the supervisory part can first establish a specific risk warning model to evaluate the potential risks of the bank's business from specific aspects,observe the flow of bank funds,point out the problems of the bank's internal risk management and correct the bank's overall governance system,which can successfully prevent The risk is successfully nipped in the bud,replacing the previous post-event supervision approach with an early prevention model.The rating of commercial banks can satisfy the general public's wish to understand the operating conditions of banks,the banks themselves can better prevent the occurrence of risks,and the regulators can better monitor the banks from specific aspects.This paper mainly refers to the "Risk Rating System for Joint Stock Commercial Banks(Provisional)" formulated by the CBRC and the theories of reputable domestic and international rating agencies such as Moody's,Standard &Poor's and Dagong International,and constructs a system of indicators that includes six categories of rating factors,namely capital adequacy,asset quality,profitability,liquidity,risk management level and social evaluation,to provide an overall rating of the current operating status of commercial banks.In this paper,the collected data are firstly subjected to missing values,normalisation and balancing.The models were constructed based on the theoretical foundations of the Xgboost,Light GBM and Catboost algorithms,and the predictions were analysed for the original dataset and the dataset optimised by the SMOTE balancing algorithm respectively,and comparing the model analysis results,it can be concluded that the SMOTE balancing algorithm can improve the model performance.Finally,based on the Stacking and Blending theories of the machine learning fusion algorithm,the fusion model was constructed and the prediction results of all the models were compared and the fusion algorithm model had a higher accuracy rate.A comparison of the model prediction results shows that among the three algorithms,Xgboost,Light GBM and Catboost,the Catboost model is more effective in predicting commercial bank risk rating problems,but is still inferior to the fusion model constructed based on Stacking and Blending theories.Further,compared with the Blending method,the Stacking fusion algorithm has higher accuracy for bank ratings,which means that in the research in the field of commercial bank risk ratings,we can focus on the machine learning model fusion algorithm to improve the accuracy of ratings for commercial banks and contribute to the risk ratings of commercial banks in China.