An Theortical And Empirical Study On Financial Crisis Early Warning Model Of Industrial Listed Enterprises Based On Oversampling Random Forest

So far,few researches have been able to fully explore the financial early warning model of industrial enterprises which account for the majority of listed companies from the actual situation of our market.The importance of industrial enterprises in social development is self-evident.Although compared with other developed countries,industrial enterprises in China start late,but the growth rate is faster,and with China's economic development into the "new normal",this rapid growth also brings a series of problems,and the transformation of industrial enterprises has become an urgent problem to be solved.First,the industrial structure of industrial enterprises is unbalanced and unreasonable,and the proportion of light industry and heavy industry in different regions is out of balance;second,due to the relatively backward start,most of the precision industrial manufacturing technology is in the hands of foreign high-tech companies,so China's industrial enterprises are in a relatively passive state;third,most of the industrial enterprises will face financing difficulties With the problem of high financing cost,China's industrial economic development has not yet realized the deep integration with financial development.In the process of continuous improvement and reform of China's science and technology and economic system,the innovation ability and level of industrial enterprises have gradually improved.However,with the promotion of internationalization and globalization of enterprises,there are also more pressures.If the enterprises cannot cope with the impact of external environment or fail to make strategic choice and transformation,it is easy for them to fall into financial difficulties.Therefore,it is an urgent problem to study the financial risk of industrial enterprises in our country,and update the sample,model and variable selection of financial early warning on the basis of previous studies to build a feasible financial distress early warning system If we can build a financial distress early warning model with high prediction accuracy,it will improve the risk prevention of enterprises,and also have reference value for investors in choosing enterprises.Therefore,this paper mainly discusses the establishment of high prediction accuracy financial early warning model from the following aspects,and according to the conclusion combined with the current situation of China's industrial enterprises,gives appropriate policy recommendations.In the aspect of sample selection,most of the previous studies choose the companies with similar scale in st and non ST companies respectively,so that the model can only be persuasive when judging these companies with specific scale,and also have certain subjectivity in the process of selection.This paper will start from the actual situation of China's market and all Samples are included to ensure that the model can be applied to different scale industrial enterprises listed companies.In terms of variable selection,traditional models use financial data to reflect corporate profitability,debt paying ability,growth ability,operation ability and cash flow indicators.In this paper,non-financial data are added,including management expenses main business income that reflect corporate management costs,the shareholding ratio of the first largest shareholder and the second largest shareholder that reflect the ownership structure Equity pledge proportion and external audit opinion.Then,by comparing the two models,we can judge whether the addition of non-financial data indicators can improve the accuracy of model prediction.For the problem of data imbalance,most of the previous studies were conducted in ST companies and non ST companies with the same number of companies.However,due to the large difference between the number of ST companies and non ST companies,the data of non ST companies can not be fully utilized.In this paper,through the way of over sampling,the sample size of ST company is increased to the same as that of non ST company,which can make more effective use of all the information and make the model more perfect.In terms of model selection,this paper considers the advantages of random forest robustness,and does not assume a specific form of regression model,so it can deal with nonlinear regression data;it can deal with very high-dimensional data,and does not need to make feature selection,has strong adaptability to data sets,does not need to standardize data sets,and has simple implementation,high accuracy and strong anti over fitting ability.