Research On Financial Distress Of Public Companies Based On Clustering Algorithm-take Manufacturing For Example

In the context of the global economy and the complex market environment, it is very important to predict the company's financial crisis for regulators, investors, creditors at the first time. So it is necessary to take financial early warning of enterprises for facilitating the listing of the information users and the company's own managers to understand the business's current financial situation, take measures in advance.Many domestic and foreign literature on financial early warning financial indicators have proven to significantly reflect the financial position, so the financial indicators used to predict whether a company into financial crisis usually. However, no one of the prediction method is perfect, there is always some disadvantages, such as the arbitrary choice of predictor variables, the correlation between variables, etc. Moreover in recent years, innovation in approach is also very limited, just improve on the technical details, so this paper tries to avoid the current financial crisis as early warning of these problems exist and to be satisfied through empirical testing the predictive effect,a fuzzy clustering algorithm of the financial crisis early warning models was proposed.In this paper, fuzzy clustering of the financial early warning methods was used. And collect the financial crisis occurred in 2009,2010 Chinese listed manufacturing companies first 3 years of data, but also selected according to the principle of matching the same number of normal financial company data, The training data and the testing data of the model are the Chinese listed companies which fall in financial crisis in 2009 and 2010.The main ideas of the fuzzy clustering include five aspects:Firstly, the paper used T-test and correlation-test statistics on the financial indicators screening; Secondly, because of rough set data only for discrete reduction, the financial data need discritized; Thirdly, the rough set attribute reduction algorithm financial indicators for further screening to determine the variables into the model after the data discritized; Fourthly,the variables selected were normalized; Finally, use netting method to cluster the listed companies, and analysis clustering results. To facilitate the identification predicted results of the model, Fisher discriminant method is selected as the compariso. In the same case of financial information, the empirical results show that the predictive of the model was better than Fisher Discriminant Model. The correct classification rate reached 85.3%,and achieved a good prediction. But this study still has some shortcomings, for example, the selection of indicators do not take into account the competitiveness of enterprises, the country's economic policy and other non-financial.And Netting method has the defect of calculating too large.Therefore, the use of fuzzy clustering to predict the listed company's financial condition is feasible, just the specific fuzzy clustering method could be improved.