| Since the 1980s,research on the probability of default has always been a hotspot for both theoretical academics and industry experts.However,many models which are widely used abroad lack the necessary implementation environment and hardware foundation in China.In order to provide practical suggestions for Chinese listed enterprises,and to provide references for commercial banks to avoid credit risks and adjust credit granting strategies,this thesis,based on the outstanding achievements in the previous studies,detailedly summarizes the research&application status of default probability models,and builds two different models using real financial data.In this thesis we reviewed domestic&foreign research status of default probability models thoroughly,analyzed their basic theories and applications.Based on the practical feasibility and applicability,the Logistic model and Cox model were selected for further research.The empirical analysis was started by dividing sample data into two groups.After filtering variables of the modeling group,we modeled the Logistic model and the Cox model separately,and adopted a unified cut-off point(threshold)to perform discriminant analysis.We used ROC&K-S curves to compare model capabilities,and obtained the conclusion that Cox model was clearly better applied in identifying both ST and non-ST companies using these data.After that,we used data of the control group for model fitting and compared these two types of models under the same conditions,which also verified the view that the Cox model performed better overall,and pointed out that the dynamic predicting ability of the Cox model allowed it to predict the survival probability of listed companies at each time-point during the survival time more accurately.Finally,combined with theoretical knowledge and research application,we put forward practical suggestions from different perspectives,summarized the shortcomings of this thesis and looked forward to further studying. |
PDF Full Text Request