| Energy is an important material basis of human social life and plays an important supporting role in the process of industrialization.In the case of limited resource storage,rapid economic growth usually leads to the shortage of energy resources,with the expansion and deepening of the global economic scale and integration,the global energy consumption is gradually increasing,which makes the international competition for energy intensified.Influenced by macro policies such as de-leveraging and supplyside structural adjustment,energy enterprises with special status in China’s economic development need to gain a foothold in the dynamic energy game competition and develop continuously,they need to pay more attention to the risk control of enterprises.The financial crisis is usually the beginning of the crisis of enterprises,financial crisis early warning system can play a warning role in the arrival of the crisis,so the establishment of energy enterprise financial risk early warning model is of great significance.Based on the above analysis,this paper takes China’s coal,oil,natural gas,electricity,non-ferrous metals and other 5 industries of energy listed companies as the research object.First of all,to pick ST as the definition of health enterprises and financial crisis enterprise standards of the state indicators,secondly select 39 financial and non-financial characteristics of indicators,and successively use normal testing,correlation test and step-by-step analysis to filter out the enterprise’s financial situation has a significant contribution to the 7 characteristic indicators,to complete the establishment of the early warning model index system.Due to the nonlinear relationship between the energy market,the nonlinear support vector machine SVM artificial intelligence early warning model is introduced,and due to the extreme imbalance of the sample size of crisis enterprises and healthy enterprises,this paper uses mixed sampling ODR-ADASYN for the treatment of non-equilibrium samples.Establish the ODR-ADASYN-SVM financial risk warning model.The main contents of the experiment are as follows:(1)The selection of the model core function is studied.The construction of nuclear function is an important way to realize the nonlinear classification of support vector machine,and the choice of nuclear function has no doubt about the predictive performance of the classifier.In this paper,four nuclear functions,such as Linear,polynomial,RBF,and Sigmoid,which are currently representative,are selected for comparative analysis.The study found that the model forecast performance is different for different prediction periods.Among them,the T-1 period selects polynomial nuclear function,selects Sigmoid nuclear function in T-2 period,selects the Sigmoid nuclear function in T-3 period,and makes the best prediction.(2)Analysis of the parameters of the model.The selection of model parameters is very important to the effect of the model.For a more in-depth and objective analysis of the prediction effect of the ODR-ADASYN-SVM early warning model.Taking the T-2 period as an example,this paper analyzes the predictive performance of the model from the non-equilibrium expectation level and the change of the near-neighbor sample.The results of the study found that Before 0.9,the model g value and F value stable slow growth,and at 0.9,the model G value and F value reached the highest value,when the model was 1,that is,the model degenerated to ODR-SVM model,the evaluation value suddenly minimized,indicating that simply using under sampling to sample feed equilibrium may be a large number of useful information,and can not effectively solve the SVM imbalance The sample is insufficient for two classifications.Secondly,when0.9 and 3,the model’s G、F and Acc values are up to the highest,in view of this,in the T-2 period,the above two parameter values are selected to study the model.(3)Compare different early warning models.To test whether this alert model is superior to other early warning models.The results of this paper compared this model with SVM,Logistic,Bayes and BP neural networks show that the ODR-ADASYNSVM early warning model has more stable performance in the time span and is significantly better than other early warning models in prediction accuracy.Based on the above analysis,the energy enterprise financial risk early warning model based on ODR-ADASYN-SVM has significant advantages in stability and prediction effect,which can be an effective risk prevention and response tool.In the case of enterprise management,it can be used as a tool for the prevention and control of financial risks of enterprises,to prepare for the upcoming financial risks and to introduce relevant countermeasures in a timely manner;Finally,this paper also combines the characteristics of energy-based enterprises,respectively,from the enterprise internal control,financial risk management,as well as the establishment of information technology to put forward policy advice. |
