Research On Default Probability Using Two-stage Model Of Weights Fuzzy C-means Clustering

Until May 2016,8 entities in Public Offering Bond market actually have broken their promise which results in the whole debts piled up to 10 billion RMB.Accompanying with the enormous amount of non-performing loans fully burdened by commercial banks,this instability has deeply hampered the benign development of the national finance system.For this reason,the authority has presented several measures to reverse this current by allowing assets sercuritization and encouraging debts-to-equity swapping aiming to shave off the foundation of non-performing loans ratio.However,except form these,enhancing the banks’ ability of default risk predication may substantially woks.Combining traditional measurement of default risk models by some researchers and relevant behavioral theories,this paper has listed the merits and demerits between Logistic model and Logistic model based on factor analysis.For one part,we have constructed a hybrid two-stage predicting model by fuzzy-clustering method considering two important roles exiting in unsystematic risks seriously : uncertainty and individual difference.For another,we also highlight the effects of risk weighting variables when it comes to figure out the disparities of different impact f actors especially at the stage of C-Means clustering analysis.Empirically,this paper has compared the disadvantages and advantages between hybrid model and logistic model predicting risks by using the data of listed companies in manufacturing industry fr om the Shanghai stock exchange.The results show that(1)the hybrid model will be superior to Logistic model in the accuracy of forecasting default risks;(2)notably,to predict type I statistical error,this accuracy will be strengthen heavily if the de fault threshold is set at 0.5;(3)furthermore,both solvency ability and profitability ability significantly make up the two key factors when to analysis the ratios which definitely applies to different borrowers.In sum,taking individual difference int o consideration,the hybrid model for credit default prediction proposed in this paper will promote the optimization of credit risk rating system in commercial banks,and ultimately contribute to the effective work of reducing of default risks and the pric ing of credit loans.