The Financial Early-warning Study Of Listed Companies Based On KMV Model And Support Vector Machine

The financial position of the business has always been the focus of all stakeholders, including business owners, shareholders, etc. Good financial position will encourage business owners work harder and attract more investment. Otherwise financial crisis will bring huge losses to shareholders, business owners and other interest-relating owners. To some extent, financial crisis will lead to bankrupt. In a summary, the status of financial is very important to a corporation and have a influential impact on business.The purpose of building financial distress forewarning model is to help companies to take appropriate measures in accordance with the early warning signals. Meanwhile it also have important guiding significance to assist the national securities regulatory authorities on monitoring the quality of listed companies and reducing market risk.In this article,we adopt SVM, Support Vector Machine, to build financial distress forewarning model.SVM is very stable and can overcome the’curse of dimensionality’and have other advantages over multiple linear discriminant model, linear probability model, multivariate logistic regression models.Firstly, we do some explanation to the selected indicators which can reflect the company’s profitability, solvency, growth capacity and operational capacity.They are:return on net assets, total assets net profit margin, sales net profit margin, gross margin, current ratio, quick ratio, equity ratio, inventory turnover, accounts receivable turnover ratio, total asset turnover, revenue growth, net profit margin growth, the growth rate of total assets and net assets growth rate. These indicators generally cover the financial aspects of an enterprise’s performanceSecondly, we introduce the principle of SVM, KMV, PCA, Neural Network Model and DEA. To build the financial distress forewarning model we calculate the distance to default,TE by Matlab and DEAP2.1respectively. Finally we use these two data and the financial indicators to building the model. The correctly predicted probability of this model is over80%.Compared to other research which is done based on SVM model, this article add innovatively default distance&TE during the process of building the model...