Research On Financial Crisis Early Warning Of GEM Companies Based On Random Forest

Shenzhen stock exchange established gem market in 2009.By the end of 2017,there were 710 companies listed on gem,an increase of 140 over 2016,with a year-on-year growth of 24.56%.The total market value of listed companies was 5128.8 billion yuan,96.5 billion yuan less than 2016,and 1.85% less than 2016.After nearly ten years of development,chinext has been playing an increasingly important role in China's capital market.The high growth of chinext coexists with its high risk.Therefore,it is extremely urgent to establish an effective financial risk warning model.At present,most of the studies on financial risk warning of listed companies by Chinese scholars focus on the a-share main board market with relatively stable operation,strong profitability and over 20 years of development.In order to establish a gem of the company's financial crisis early warning model,the author from 2016 to 2017 the gem listed company financial data as the breakthrough point,the comprehensive consideration of 2015-2017 data,refer to the previous classification method,with two straight losses of the semi-annual or net asset growth rate is negative for crisis classification basis,the gem companies listed before 2015 can be divided into crisis and health company two groups.Through literature review,the early warning system for financial risk at home and abroad on the collected data filtering and sorting of the early work,in the empirical part of the random forest algorithm discriminant analysis was studied for the sample of the company's financial condition,and the corresponding early warning detection model accuracy,the study found that the default parameters of random forest classification rate is not high,so in this paper,the main parameters of random forests and imbalance is optimized.In order to more intuitive reflect the random forest on gem listed company financial risk early warning problems early warning function,using the test data,the optimized model and random forest classification and regression tree(CART),support vector machine(SVM)model,this paper compares and analyzes the stability after parameter optimization of random forest model performance higher,judging more accurate conclusions.Through the random forest model,the early-warning accuracy rate of gem companies in the year before the financial crisis can reach more than 84%.Overall,based on the knowledge of the financial statistics,finance,using random forest algorithm is widely used in the recent gem listed company financial crisis forecast,early warning model is established,through the crisis early warning model to the gem listed companies may be data to predict financial crisis of the previous year,the gem listed company's financial risk management has a certain reference value.