Research On Systemic Financial Risk Measurement Of Commercial Banks In China

In recent years,the information revolution represented by new technologies such as big data,cloud computing,artificial intelligence and blockchain has not only promoted the reform and development of China’s traditional finance,especially commercial banks,but also promoted the transformation of the old and new driving forces of China’s economy.However,the system of China’s financial system is still not perfect,and the rapid development of finance has caused the accumulation of risks.In addition,all kinds of derivative risks caused by COVID-19 impact can not be ignored.Therefore,how to measure and prevent financial risks is much concerned by the government.The systemic financial risk of commercial banks is an important part of China’s financial risk,the focus of financial risk prevention,and one of the research hotspots of scholars at home and abroad.Traditional systematic financial risk measurement methods can describe the fluctuation of financial market and estimate the future risk situation,such as Covar method and financial pressure index method,but ignore the problem of how the risk is transmitted layer by layer and eventually leads to risk loss.To solve this problem,this paper proposes an adaptive weighted Bayesian network model,which can not only predict the future risk situation,but also reflect the causal relationship between the indicators of commercial banks.Random forest is an integrated learning method to judge the importance of features by calculating the classification error of out of bag data.Therefore,this paper selects the data from the first quarter of 2019 to the first quarter of 2021,and studies the measurement of systemic financial risk of commercial banks combined with random forest and adaptive weighted Bayesian network.Firstly,due to the characteristics of high dimension and high complexity of bank special data,Bayesian network has some difficulties in learning the complex network structure with multiple nodes.Therefore,in order to solve this problem,based on the different periods of the epidemic,this paper uses the rise and fall of bank stock closing price to reflect the risk,constructs an index system that has a significant impact on the systematic financial risk of commercial banks in different periods of the epidemic,realizes the dimensionality reduction of the data,and dynamically analyzes the significant indicators in each period of the epidemic.The results show that the cost income ratio and the deposit loan ratio,both of which have a significant impact on the systemic financial risk of banks in the economic stability period,the epidemic impact period and the post epidemic period,need to be paid close attention to by commercial banks.Secondly,learning the Bayesian network structure based on MATLAB software,this paper analyzes which factors affect the level of systemic financial risk of commercial banks,the joint influence degree of these factors,and how the risk is generated and transmitted layer by layer in the process of bank operation,resulting in risk loss.This paper uses CACC algorithm to discretize the continuous data of banks,then uses MCMC algorithm to learn the network structure,constructs the Bayesian network of systemic financial risk of commercial banks in different periods of the epidemic,and analyzes the influence relationship between each index.Experiments show that the systemic financial risk in the period of economic stability is mainly affected by the ratio of non-performing loans and the ratio of cost to income;The systemic financial risk during the impact period of the epidemic is mainly affected by the cost income ratio,net interest margin and loan loss reserve adequacy ratio;The systemic financial risk in the post epidemic period is mainly affected by the provision coverage of non-performing loans and non interest bearing liabilities.Thirdly,starting from the types of different commercial banks and selecting different feature combinations,the risk prediction results of Bayesian network are different.Therefore,starting from the types of commercial banks and based on the classification performance of different feature combinations,this paper establishes three classifiers and predicts the risk level of commercial banks in the next quarter.Finally,the weighted decision-making value of the financial system is obtained through the weighted decision-making of the bank.Experiments show that the adaptive weighted Bayesian network has a great improvement in risk prediction compared with the Bayesian network with a single classifier.Finally,the research results of this paper are summarized,and the shortcomings of this paper and the prospect of possible future research are given.