Research On Green Credit Risk Assessment Based On SMOTE-LightGBM Algorith

In recent years,China’s green credit policy has encouraged banking financial institutions to vigorously develop green credit,and in order to actively implement the policy requirements,banks are constantly exploring green credit implementation standards and evaluation systems to improve the efficiency and implementation effect of green credit business review.Since green credit is different from traditional credit,the risks faced by green credit business are more complex,so how banks should accurately and efficiently assess the risks of green credit projects has become the primary issue for banks to issue green loans.Therefore,this paper constructs a green credit credit risk assessment index system for listed companies and establishes a green credit credit risk assessment model for listed companies based on the SMOTELightGBM algorithm,which is a green credit risk assessment index system The improvement and application of green credit credit risk assessment methods provide a basis.Based on the basic information,financial data and green credit data of listed companies in China from 2018 to 2020,a preliminary index system for green credit credit risk assessment of listed companies was established,and then the corresponding company classification data and SMOTE-LightGBM model were screened for characteristics According to the importance score of indicators obtained by LightGBM algorithm,28 indicators with high importance degree were screened out,and a green credit credit risk assessment index system for listed companies was constructed.Then,the SMOTE-LightGBM algorithm is used to establish a green credit credit risk assessment model for listed companies,and the appropriate data processing method and training set test set division are selected according to the empirical analysis,and the missing value processing,min-max standardization,and SMOTE oversampling processing are carried out on the data set,and the data set is processed 75% of the data is used as the training set and 25% of the data is used as the test set for empirical research,and the empirical results are obtained through the parameter optimization of the LightGBM model,and then the applicability of the model is analyzed from the evaluation indicators such as the confusion matrix,the evaluation report of the classification model and the ROC curve.Then,based on KNN,decision tree,random forest,GBDT and XGBoost algorithms,the green credit risk assessment model of listed companies is established,and the test set is divided according to the same data processing method and training set test set as before,and passed by Python The program conducts empirical analysis,compares and analyzes the quality of each model through the results of evaluation indicators such as confusion matrix,accuracy,precision,recall,f1＿score,ROC curve and AUC value of each model,combined with the analysis results of LightGBM model.At the same time,according to the characteristics of the LightGBM model,the importance of each index is obtained,which is conducive to improving the audit indicators of the bank’s green credit business and the company’s timely industrial adjustment decision.Finally,conclusions are drawn and relevant recommendations are made.In order to promote the sustainable development of green credit in China,it is necessary to establish and improve the green credit credit risk assessment system,and strengthen green credit risk management from three aspects: government,banks and companies.