Chinese Listed Companies’Credit Risk Early-Worning Based On Bayesian Network

Since August2007, United States subprime mortgage crisis triggered a financial crisis sweeping the world, sharp turbulence in the financial markets, which suffered from tired. However, one after another,2009European debt crisis to make the European economy into a quagmire, has cast a shadow over economic recovery. The European debt crisis is a United States continuation and deepening of the subprime mortgage crisis, the root cause is government debt exceeds its range as a result of the credit risk. Therefore, the credit risk of the national debt will have disastrous effects on the economy. Similarly, the credit risk of enterprises will also have a very serious problem. Early warning effectively of enterprises’credit risk becomes a problem worthy of in-depth study.This paper, combined of Bayesian network theory, theory of financial early-warning and MATLAB programming techniques and SPSS statistical analysis techniques to include multivariate analysis, artificial intelligence, data mining and other previous research experience, will establish a model based on Bayesian network for credit risk early-warning for listed companies. At the same time, this study describes the construction of models for large samples, multivariate analysis, showed that both models have better predictive accuracy and stability, and made a preliminary exploration and practice on model.First of all, research methods and theory of carding people. Through analysis, identify the internal factors of enterprises’ credit risk. According to studies on selected standard samples and financial indicators, preliminary15financial indicators were selected as models of alternative indicators and alternative indicators for analysis and verification, a P-P normal distribution test indicators individually, Person correlation and t-test. Secondly, systematic introduction of Bayesian network modeling process--structure and parameter learning. Focused on two of the Institute’s two kinds of Bayesian network structure learning algorithms (K2and MCMC algorithms) and two parameter learning algorithms for Bayesian network (maximum likelihood estimation and maximum a posteriori probability). Finally, combined with MATLAB programming sample set and index system of listed company based on Bayesian network and empirical analysis of credit risk early-warning models. This study also analysis and discuss the application of the model.This study on enterprise credit risk research on theory and model of enterprises’ early-warning is the inheritance and development of past research. Credit risk early-warning model based on Bayesian Networks have developed credit risk early-warning method of enterprise. The model can provide theoretical guidance and decision support for enterprises’managers, bankers, investors, regulators and other stakeholders. It is of great theoretical significance and practical significance.