The Study On Financial Risk Forewarning Of Listed Companies Based On Data Mining

With the rapid development of our country economy,the extent of damage on listed companies made by financial risk is deepen increasingly,it not only affect the existence and development of company,but also will bring enormous loss to related benefits.Therefore,it is imperative to resolve the problem for our country listed company that how to mine the information which can forecast the financial risk.But the gradual mature of data mining technique resolve the problem at right moment.Data mining have the function to mine potential rules and knowledge from amounts of historical data.So the usage of data mining technique to carry on the research to the financial risk forewarning of our country listed companies and setting up an financial risk forewarning model that apply effectively,have very important of realistic meaning and theories meaning.The main research of this paper is the application of the data mining's logistic regression, neural network and decision tree in the financial risk forewarning.Firstly,related theory of financial risk forewarning and data mining technique are introduced.Secondly,research process of financial risk forewarning modle is constructed,and a detailed description of the three data mining technique used is made.Then,considering our country listed company characteristics,the financial risk forewarning index system is constructed.Finally,77 special treated manufacturing listed companies form 2006 to 2007 and their 77 corresponding manufacturing listed companies which are informal financial conditions are selected as the comparative samples,based on financial data ahead of 2-4 year of when each company is special treated,contrasted by the CRISP-DM framework.The data mining software Clementine is used to the realization of application of data mining in the financial risk forewarning.From the research result,financial risk forewarning modle constructed by data mining have stong forewarning ability,From the forecast precision of three models made in the paper is higher when enters the time when the company is special treated nearer.Three methods have better forecast in the short time forewarning,but the neural network and decision tree is better than logistic regression based on traditions statistics analysis in the long time forewarning.Data mining based on knowledge discover is not only suitable for the shot time forewaring,but also for the long time.So data mining is applied absolutely to the financial risk forewarning, and attain to support the decision.