Research On Credit Risk Early Warning Of Listed Energy Companies Based On The FOA-BP Model

For a long time,the energy industry has played a fundamental role in China’s economic development.Its development status is not only related to China’s energy security,but also closely related to China’s people’s livelihood issues.It is an indispensable and important member of the real economy.However,since China entered the 21 st century,with the change of energy consumption structure,the consumption of traditional energy has shifted from relying mainly on coal to oil and natural gas.The lack of oil and natural gas resources in China has led to a high degree of external dependence on oil and other energy sources.Especially,in the With the advocacy of the "Belt and Road Initiative",the traditional energy security has been affected due to the political security of the countries along the route.However,new energy faces difficulties in technology development,high R & D cost,large capital demand,and long investment cycle internally,and facing industry competition pressure and international trade protection externally.the dual role of internal and external factors will make new energy companies have high credit risks.Only by establishing a scientific and effective early warning model of credit risk for listed energy companies,can we accurately predict the credit risk they face,so as to effectively prevent and respond to the credit risk,and achieve the purpose of maintaining energy security,and promoting social and economic harmony and stability.Up to now,many scholars and researchers have used a variety of early warning models to predict credit risk,especially,the BP neural network model with superior prediction performance,which is more popular in the prediction of credit risk.However,considering that the selection of characteristic indicators has a very important influence on the traditional BP model for predicting the credit risk.When the characteristic indicators used by the risk research institute are extracted,quite a large number of scholars use the extraction based on the experience of previous literature methods and component analysis process feature indicators.These two methods mainly integrate feature indicators rather than reduce them,resulting in lack of interpretability of feature indicators with traditional BP neural network models.The setting of thresholds and weights is random,so it also directly affects the performance of the model on credit risk early warning.In order to solve these problems,this article uses indexof Lasso screening method to extract the characteristics of credit risk indicators,and quotes FOA The parameter optimization method optimizes theparameters of the traditional BP model.Only some parameters that control the FOA algorithm are used to achieve optimization,which overcomes the shortcomings of the BP model,such as slow training speed,difficulty in convergence,and easy to fall into local minimums.An improved BP model that is effective for the prediction of credit risk of Chinese energy listed companies,namely FOA-BP credit risk Risk early warning model,and then through empirical research to verify the superiority of the FOA-BP model in credit risk early warning performance.Empirical research shows:1.The paper solved the problem of extracting credit risk characteristic indicators.In this paper,after standardizing and normalizing the 36 selected credit risk characteristic indicators,processing method of Lasso is introduced to extract the characteristic indicators,and 20 credit risk characteristic indicators are selected,which can improve the model’s credit risk to the accuracy of early warning in a certain extent.2.The paper solved the problem of parameter optimization.The FOA algorithm is used to replace the gradient solution method,but the parameters of the BP neural network credit risk early warning model are optimized in an iterative manner.And the FOA-BP model is constructed to achieve early warning of credit risk.Empirical research shows that it is different from its traditional BP neural network early warning model.In comparison,the improved model achieves higher convergence efficiency and stronger adaptability.Under the different sample ratios of 7: 3,6: 4,and 5: 5,the BP neural network model modified by FOA is more generalized in predicting the credit risk of listed energy companies in China3.In the prediction experiment of the traditional BP neural network credit risk early warning model,we can know that under the different sample ratios of 7: 3,6: 4,and 5: 5,although the ACC value of the prediction result of the traditional BP neural network model is high,there are more serious errors of the second type,which indicates that it is inaccurate to predict the credit risk simply by using the BP neural network credit risk early warning model.4.From the comparative study of credit risk prediction performance of SVM model based on FOA algorithm,Logit model based on FOA algorithm and FOA-BP model,at the different sample ratios such as: 7: 3,6: 4 and 5: 5,the credit risk prediction performance of SVM model and Logit model based on FOA algorithm have some features,which the prediction accuracy value,G value,and F value are not as good as the FOA-BP model.The FOA-BP model is more generalized for early warning of credit risk.Through the above research,this paper believes that the FOA-BP credit risk early warning model based on FOA to improve BP neural network parameters is an effective operation tool for investors,operators,and regulators to prevent and respond to credit risk.For investors,you can use the FOA-BP credit risk early warning model to capture credit risk signals in advance,optimize investment strategies,and reduceinvestment losses due to credit risk.For operators,you can use the FOA-BP credit risk model effectively predict credit risks in advance,strengthen business management,and formulate and implement credit risk response plans to achieve healthy and stable development of enterprises.For regulators,it is possible to strengthen capital market credit risk prevention and control measures to pre-establish reasonable and effective credit risk policy solving measures to reduce the impact of credit risk on the capital market.