Study On Biological Medicine Listed Companies' Financial Distress Based On Support Vector Machine

With the increasing pressure of market competition,the status of listed companies being special treated frequently due to the financial crisis.With the bio-pharmaceutical industry is listed as our strategic high-tech industries,China's bio-pharmaceutical enterprises get rapid development and expansion.In recent years,rapid expansion of China's capital market and constantly improvement of the market system,the competition between the industry have become more and more intense,companies are always faced being with eliminated in the threat of market.Therefore,on the basis of public real data,how to build an effective financial crisis warning model have practical significance.In the first chapter we have discussed the background of the financial crisis and proposed research framework and methodology.A Summary of theories in the second chapter after the introduction of the financial crisis early warning system describes the theoretical scholars reviewed the definition of the financial crisis,the impact factors and the modeling method.According to the model the research method selected,the system introduces the SVM theory,namely optimization theory and convex duality theory,linear and nonlinear problems as well as problems of several common kernel function formula.Finally,the support vector machine as the popular applications and research in the field of data mining methods in the recent years.On this basis,it is determined to support the establishment thinking vector machine,firstly selecting the bio-pharmaceutical companies in 129 normal companies and 43 ST companies in the sample,filter characteristics based on the principles and bio-pharmaceutical companies in the literature selected indicators of industry characteristics indicators,from a financial point of view illustrates the inner meaning of each index.After selecting data and financial indicators,in order to avoid differences in results caused by different dimension of the data is normalized.To overcome multicollinearity between financial indicators using principal component analysis.Kernel function in support vector machines and kernel of choice,combined with domestic and foreign scholars Reference R software package e107 choosing and setting parameters on nuclear function,namely the financial crisis early warning model established four kinds of kernel function test and obtain test results.The study concluded that four kinds of kernel stability and accuracy of financial crisis warning varies.Wherein the warning accuracy radial kernel function of the highest polynomial kernel function and Sigmoid kernel function prediction accuracy rate,followed by linear kernel warning less than ideal.While the use of financial crisis early warning theory classical Logistic regression analysis to predict,by contrast,showed Logistic model training set accuracy better than SVM,but the prediction accuracy is much lower than SVM.By analyzing the results of the above-described sorting and empirical theory of literature,I believe that support vector machine model for biomedical listed companies'financial early warning of practical significance,can provide a reference value for our investors,regulators and the company's management.Finally,the prospect of the support vector machine in the financial crisis early warning of further in-depth study.