Research On Financial Crisis Early Warning Of Listed Companies Based On TPDA-NB

Since 2007, the US subprime mortgage crisis triggered by the financial crisis sweeping the globe, making the world’s financial markets turbulent, the economy is deeply tired also. However, the 2009 outbreak of the debt crisis dragged into the quagmire of the global economy once again. This not only makes the risk of a sharp rise in the market, but also makes risk management face more severe challenges. Therefore, how to strengthen the awareness of risk management and improve the accuracy of risk prediction, maintain harmonious and stable economy and society, not only the main task facing the government economic management departments, it has also been attracted a lot of attention in academia.It is noteworthy that, with the prosperity and development of China’s securities market, more and more companies to obtain funds through the listing to expand the development of listed companies has become the core strength of China’s economic development. In addition to that, the financial crisis within the enterprise group is contagious lead to further government economic management departments and other stakeholders and investors attention of Chinese listed companies’ financial crisis. Listed companies in the event of credit defaults, investors will not only suffer from huge losses, and may even lead to serious consequences of bankruptcy and social unrest. Therefore, building a scientific and effective financial crisis prediction method has an important practical significance.Based on this, this paper takes China’s listing corporation as the research object firstly, using the normality test, test parameters and nonparametric test and multiple linear feature extraction index can significantly describe the financial crisis of the listing corporation; Furthermore, for the sake of overcoming the Bayesian network(BN)’s limitations of relying too much on the sample data when learning the initial network structure, the naive Bayesian network model(NB) will be regarded as the initial network structure. And then introduce a three-phase dependency analysis(TPDA) algorithm based on constraints to overcome the conditional independence assumption which is relied too much on. Then, constructing an improved Bayesian network model, which is TPDA-NB model, to warn the financial crisis of listed companies. Finally, testing set is applied to compare TPDA-NB model with NB model, Logistic model, Neural Network Model, and, the paired sample T test is also adopted to analyze the four models’ prediction accuracy differences.Empirical study results show that, firstly, by comparison Logistic model, NB model with TPDA-NB model, it is found that, in the forecasting accuracy and stability, the significant differences not only between NB and Logistic model, but also exist between logistic model and TPDA-NB model, which are more significant; secondly, by comparison the neural network model, NB model with TPDA-NB model. Finding that, between and TPDA-NB model exist significant differences in forecasting precision and stability, but the differences between neural network model and NB model are weaker; Finally, which is more importantly, TPDA-NB model can effectively enhance the NB model on the prediction accuracy and stability of the financial crisis of listed companies.The results of investigation illustrate that, TPDA-NB model can be used to predict the China’s listed companies’ financial crisis accurately, which has an broad application prospects in the field of risk management. For investors, could use TPDA-NB model in advance to capture risk signal, and then make rational investment decisions to avoid risks brought about the loss; for government economic management, can use TPDA-NB model of possible areas of risk prediction, then timely development of an reasonable regulatory policy, to stabilize the market order, promote the sustained and healthy development of the economy.