Build A Listed Company Financial Crisis Early-warning Model Research

In recent years, China’s stock market since the financial circumstances of any abnormalST listed companies are gradually increasing. This kind of abnormal financial position besidesbring serious damage to the interests of the company, also harm the interests of creditors,shareholders and other stakeholders, even seriously disrupted the normal order of marketenvironment. So, the listed company in making full use of publicly available data on the basisof the construction of a stable and reliable financial crisis early warning model, help to avoidfinancial crisis and bankruptcy.In this paper, through the statistical review of some of the relevant literature andcombining with the theory of knowledge in financial management initially select25financialcrisis early-warning index. Making a normal distribution test to the primary25financial crisisearly-warning indicator, financial crisis early-warning indicators to obey normal distributionare made ST and non ST samples T test；the early-warning indicators of the financial crisisdoes not obey normal distribution are made a nonparametric tests of ST and non ST samples,according to the results of T test and nonparametric tests, excluding significant financial crisisearly-warning indicators, finally determining the21early-warning indicators of the financialcrisis, the financial index of20, non financial index1. In this paper, making a KMO test(correlation test) to20financial indicators, the test results show that there is a correlationbetween the financial index. In order to overcome the correlation between financial indicators,so using the principal component analysis (pca) to extract eight principal component toreplace the original20financial indicators, and the eight principal component and a directgeneration of non-financial indicators into the Logistic regression, were analyzed. Accordingto the results of Logistic regression analysis, and removal of variables is not obvious, furtherconcluded that the Logistic financial crisis warning model.The samples are choosed only from A shares in the listing Corporation of the point oftime to market a distance of more than5years, with the sample of ST and non ST state as asign of whether the company falls into the financial crisis，according to the principle ofrandom sampling selects116companies which are2010to2013for the first time by in the STas the financial crisis of the company; according to the1:1industry matching principle andselects116companies which have never been ST A shares of listing Corporation as thefinancial health of the company sample. Among them using2010-2012years T-2years (T for the year) data to construct the early-warning model of financial crisis, using2013T-2years ofdata to test the early-warning model, test results show that the model accuracy is as high as80.8%, achieved the desired effect.