Early Warning Model Of Financial Distress In Listed Companies Based On SVM

This thesis first describes the background of the establishment of the financial distress model,then summarizes the research of scholars at home and abroad from three directions.Next,this thesis introduces the steps of selecting samples and indicators,then selects the2016 data of 2018-ST company and 2018-non-ST company from CSMAR.Then this thesis selects two sets of indicators according to the steps.The first set is indicators that does not deal with multicollinearity,and the second set is indicators obtained by principal component analysis.Then,this thesis introduces the SVM model,and finally selects the Gaussian kernel function SVM model as our early warning model,and obtains the prediction results of two sets of indicators respectively.The experimental results in this thesis show that after combining the two sets of indicators with the Gaussian kernel function SVM model,they can get good prediction results,but the prediction effect of the first set of indicators is better than the second set of indicators.There are two possible reasons for this result.The first point is that after the principal component analysis,the second set of indicators lost a lot of information.The second point is that the Gaussian kernel function SVM model is not sensitive to the correlation between indicator variables.