Early-warning For Small Medium Enterprise Financial Crisis Based On Logistic Regression

In the thesis, the membership companies listed in one of Beijing small medium enterprises association are treated as research sample, and 22 variables are picked from financial ratios, which could play significant part in Logistic Regression Model to forecast the company's financial crisis in one year early.Firstly, the initial model variables contain 22 financial ratios which has been proved to be effect in forecasting model in previous literature study. Selecting the samples by random, samples are divided into the training group and the forecasting group, and the number of companies with financial crisis equals the one with financial normal in training group. To constrain the effect of multi-co linearity, the normality test and relevance test are performance. With the logistic regression, the right percentage of forecast is 79.0%, and the false rate is 21.0%.Secondly, to improving the accuracy of the model, cluster analysis and factor analysis are put into use. With cluster analysis, the samples with the same financial characteristic are put together. After the factor analysis, the 22 variables are reduced to 7. The precision rate of the logistic model raises to 84.76%,while the false rate down to 15.24%.This research finds out that when analyzing the SME model with less industry information, this method can be effective. Using between-groups linkage cluster and factor analysis could improve the precision rate.Finally, the paper discussed applying advantage, the problems which should be noticed and the defect of the model.