Research On Prediction Of Large And Medium-sized Coal Enterprises Based On Machine Learning

The coal industry is an industry sector in China’s coal resources exploration,mining,production,storage and transportation,processing,environmental protection and other aspects.It has played an important role in the global economic development for a long time.Coal is not only the main energy of China’s primary energy,but also the basic energy and important raw materials.As an important part of primary energy,it plays an important role in the stable development of the national economy.With the improvement of the marketization degree of China’s coal industry,the competition among coal enterprises is becoming increasingly fierce.Coal enterprises will encounter credit risks caused by various factors at any time in the process of production and operation.Therefore,in order to find the risks as soon as possible and make corresponding adjustments,it is necessary to give timely early warning of credit risks.In this context,according to the relevant financial and non-financial data disclosed by large and medium-sized coal enterprises,this paper studies the credit risk early warning methods and mechanisms of large and medium-sized coal enterprises.The specific research contents include:(1)Analyze the main influencing factors of credit risk of large and medium-sized coal enterprises.Through extensive combing of relevant literature at home and abroad,according to the "5C principle" generally recognized by the industry,this paper comprehensively considers the risk characteristics of large and medium-sized coal enterprises from the external environment,wealth creation ability and debt repayment sources of coal enterprises,creatively puts forward new indicators reflecting the credit risk of coal enterprises,and establishes an index system.(2)The feature selection method of Filter-Wrapper framework is proposed to screen the credit risk index system of large and medium-sized coal enterprises.In order to effectively distinguish the significant characteristics of default and non default States,the feature selection method of Filter and Wrapper selection algorithm is proposed.Firstly,the source of credit risk data of large and medium-sized coal enterprises is explained,and then the construction and training process of the model are studied,which lays a solid foundation for the follow-up credit risk early warning model.(3)Research on credit risk prediction based on Stacking model with improved unbalanced data.Aiming at the imbalance of data in the optimal index system,the improved Borderline SMOTE-2 algorithm is used for sampling processing.Secondly,an integrated model is constructed to fuse and predict the multi-dimensional data to form a credit risk early warning model for large and medium-sized coal enterprises,which is applied to the actual data of coal enterprises.(4)Example application of coal enterprises.The empirical analysis results of Listed Companies in the coal industry show that the Filter-Wrapper screening algorithm can effectively identify the default status of enterprises and better reflect the characteristics of coal enterprises.Through the comparative analysis with the existing classification models,it can be seen that the early warning model established in this paper has a more accurate and stable classification effect on the credit risk assessment of large and medium-sized coal enterprises,and has a stronger ability to identify default samples.It can be applied to the credit risk prediction of large and medium-sized coal enterprises.