With the development of economy,the environmental problems are becoming more and more serious.The government has issued policies to support the development of green finance.In this context,how to promote the construction of ecological civilization and promote sustainable economic development has become a concern of many scholars.Green credit has become an effective way to promote sustainable development because of its special environmental protection function.However,the essence of green credit is still credit,containing risks that can not be ignored.There are many problems in the field of green credit risk,such as unbalanced data,small sample,nonlinearity,uncertainty and so on.The traditional rating methods can not adapt to it.First of all,the concept of green credit,development status analysis.The development of green credit started later than that of western countries,but in recent years,with the support of national policies,the scale of green credit has developed rapidly,and has become the first place in China,green credit environment performance gradually becomes significant.Secondly,starting from the financial and non-financial aspects,the initial evaluation index system is constructed,the index is screened by the Select From Model model,and 15 indicators are retained to form the final green credit credit risk evaluation index system.Then,taking the heavily polluting companies listed on the main market of Shenzhen Stock Exchange as the empirical objects,the random forest model with default parameters,the random forest model with grid search,the random forest model with genetic algorithm,and the Gaussian kernel support vector machine model were used for early warning.The results showed that the effect of the kernel support vector machine model was not as good as the random forest model.The overall performance of the genetic algorithm random forest model is better than the other two random forest models,and its AUC value reaches 74.07%,which is 2.74%,6.18% and 2.74% higher than that of the random forest model with default parameters,the random forest model with grid search and the support vector machine model with Gaussian kernel,respectively.Finally,the stochastic forest model of genetic algorithm was used to sort the importance of the indicators,and it was concluded that the first-level indicators reflected by the indicators that had a greater impact on green credit risk were the enterprise’s environmental governance,social responsibility,profitability and debt paying ability,and policy suggestions were put forward for the four dimensions respectively. |