Research on Financial Credit Risk of Manufacturing Enterprises Under Unbalanced Data:Based on Stress Testing and Machine Learning
How to effectively assess the financial credit risk status of enterprises is the current research focus in the field of risk warning.Taking Chinese manufacturing enterprises as an example,firstly,through the principal component analysis and K-mean clustering to comprehensively score and classify the financial credit risk of enterprises,an explo-ration was made on the importance of indicators in depth;then,SMOTE oversampling was used to solve the problem of category imbalance in order to improve the prediction effect of the machine learning model;finally,an evaluation was made on the prediction effect of each machine learning model,and taking the model with outstanding performance as the stress transfer model,an analysis was made on the stress resistance of enterprises under different segments through stress testing.The study found that:1)there is a significant difference in the degree of influence of each credit indicator on the financial credit risk of manufacturing companies.For example,the most influential is the industry solvency,and the least influential is the enterprise operating ability;2)in the stress test,compared with other models,the MLP model has the best overall prediction effect,with the smallest decrease in(P)MLP and the smallest increase in CVMLP;3)with the increase of stress factors,the stress resistance curve of manufacturing enterprises under each sub-sector decreases significantly.Taking the decline as a criterion,the general equipment manufacturing enterprises have stronger stress resistance,while the specialized equipment manufacturing enterprises have smaller stress resistance.The research results can help stakeholders more effectively assess and manage the financial credit risk of manufacturing enterprises,reduce risk exposure,and promote the healthy development of enterprises.
manufacturing enterprisesfinancial credit riskstress testingmachine learningunbalanced data
万方数据
