Research On Bank Credit Rating Model Based On Integrated Learning

The banking industry,being the oldest and most significant in the financial sector,is linked to all economic activities in society.This is due to the emergence of economic globalization,which has caused economic ties to become more intimate across the globe and has also connected the financial sector globally.Once a large bank has a problem with its business situation,it often causes a butterfly effect.Regulators and Chinese banking industry alike face a formidable task: to avert commercial bank difficulties and to offer superior oversight and credit ratings for commercial banks.The West’s prior economic growth has made the financial system more ideal than its domestic counterpart,and the renowned international rating agencies boast an impeccable rating system for commercial banks.The ever-evolving market environment renders the credit rating of commercial banks in China inadequate to meet it.The market’s ability to observe the operating conditions of commercial banks is greatly aided by the continual refinement and optimization of the domestic and foreign credit rating systems for these banks.The credit rating of commercial banks can provide customers and investors in the market with a more direct insight into the bank’s operations and credit standing;furthermore,banks can reduce risks based on the credit rating results.The investors and clients can gain insight into the operations of commercial banks through their credit ratings,while the bank itself can be better equipped to avert potential risks and the regulator can be more effective in supervising the bank from certain angles.This paper mainly refers to the Risk Rating System for Joint-stock Commercial Banks formulated by the CBRC and the rating systems of well-known domestic and foreign institutions such as Standard & Poor’s and Fitch,and constructs a rating perspective including capital adequacy,asset quality,liquidity,profitability,risk management capability and social evaluation six aspects,with data from text analysis,to evaluate the overall operating conditions of commercial banks.In this paper,the collected data are first processed for missing values,data standardization and data balancing.Comparing the predictive ability of Ada Boost,XGBoost,Light GBM and Random Forest models,the credit rating models of commercial banks are constructed.It is clear that Light GBM is the more suitable model for the credit rating system of commercial banks in this paper,as evidenced by the model prediction results.Finally,the comparison of the ranking results of the importance of the characteristics of the four models in this paper shows that risk-weighted assets/total loans,operating profit,total assets and loan provisioning ratio are significant indicators in the credit rating models,and the degree of digital transformation,a non-traditional financial data,can also play a predictive function to a large extent,which means that in Consequently,in the study of commercial bank credit rating,we can strive to make a breakthrough in non-traditional financial indicators and acquire novel types of data through text analysis and other means to analyze credit rating from various angles.