Using Bayesian Network For Financial Early Warning Of The Public Companies

It is very important to forecast financial status of listed company in the new century. It also has important meaning how to establish good financial early warning model.This paper reviewed the domestic and international research results at first, summarized shortcomings and excellences of the financial warning model that scholars adopted in the past. The paper put forward Bayesian Network model on the basis of research results. According to the characteristics and experiences that Bayesian Network success in the other fields, we analyze the advantage that the Bayesian Network model will apply to the financial early warning. Recently years, many scholars have used Bayesian Network to study problems in management field. Bayesian Network has many qualities. Because those characteristics, Bayesian Network is dynamic and its variables need not accord with normal distribution. It also clearly outbalances ANN-the way of neutral networks model because the process of establishing Bayesian Network is transparent.This paper study sets up financial early warning model by using Bayesian Network. The model can improve the veracity and pertinence of prediction for listed companies. The paper introduces the fundamental of the Bayesian Network and the currency method of establishing the model. We also introduce some particular algorithm for Bayesian Network. The algorithm relate to the probability theory and computer software. In the fourth chapter, we analyze and select the sample data firstly. Secondly, we select the right financial ratios. Thirdly, to classification those financial ratios. Lastly, to ascertain causality of every nodes by computing the conditional probability. According to causality of every node, the location of nodes should be ascertained. If we have found that location of nodes disaccord to location that orientationedby expert knowledges, we should adjust the location .In the end, we set up Bayesian Network model.Utilizing listed companies data in 2003 to validate the model that the study has set up, the rate of accuracy is up to 92%. The results indicate that Bayesian Network early warning model for financial failures has the higher rate of accuracy. We also sum up excellences and shortcomings that using the Bayesian network for financial early warning and put forward some advices for the subsequent investigator.