Research On Forewarning Model Of Listed Companies Financial Distress Based On Bayesian Network

With the opening up of the financial market in China,and the development of the securities market,listed companies are playing an important role in allocation of resources,raise of funds and promotion of company innovation etc.And their impact on the whole economy is deeper.Therefore,the healthy operation of listed companies is extremely important.If large-scaleenterprises fall into financial distress,like subprime crisis,not only a single enterprise will get big loss or even bankruptcy liquidation,the whole economy and society suffer a huge concussion.Therefore,predicting the degree of the possible financial distress through establishment of mathematical model is increasingly important to investors,stakeholder,government and the society.Bayesian network is a tool for reasoning and analysis of the uncertainty by the knowledge of mathematical statistics.And it can be expressed in graphics,thus,the dependency relations among variables are visible.The financial risk is random event with its uncertainty.So,the prediction of enterprise financial risk via Bayesian networks is very convenient.Although domestic scholars have many studies on the financial distress prediction,the use of this method is very rare.Song Yao and Zhu Huiming applied Bayesian network in financial distress prediction,but its forecast model contains only financial index,and they didn't take the advantage of computer tools to build forecasting model.On the base of a large number of literature reading,this paper generalizes the definition of financial distress and its influencing factors.The theory of Bayesian network and principal component analysis are introduced.Based on the analysis of the financial distress factors,ownership structure,besides of financial index,governance structure and macro-economic factors are considered in the model.Financial indicators,non-financial indicators and macroeconomic indicators are various and co-linear,considering the accuracy and complexity of the model,so this paper selects the main factors of financial indicators,non-financial indicators,macroeconomic indicators,and calculates the comprehensive score by principal component analysis.the four nodes of the Bayesian network and the value ranges of the nodes are determined by Mann Whitney U test and principal component analysis,for instance,macroeconomic score,ownership concentration,chairman and CEO duality,as well as the financial indicators comprehensive score.According to the expert knowledge,the Bayesian network structure learning is conducted to determine the relationship among the nodes.Sixteen manufacturing companies which are special treated in 2014 and forty-eight non special-treated manufacturing listing Corporation are be selected as the total sample.And the simple is divided into the learning sample which is used to study the parameters of the model and the test sample,which is used to test the prediction accuracy of the model.The financial distress prediction model is constructed based on the Bayesian network and principal component analysis.And the results show that the accuracy of the model is 90%,so it can predict the financial distress of the company.