Research On The Financial Distress Early Warning Model Of The Chinese Listed Company Based On Neural Network

With the development of commodity economy, stock companies become the basis of the securities market. And the position and results of these companies draw the attention of investor, creditor, shareholder and manager. Chinese enterprises inevitable encounter the risk of unpredictable in the process of economic globalization, because the existence of the shortcoming in management. How to deal with financial risk is directly related to the survival of enterprises, therefore, creating the financial crisis warning model is significant to enhance the management level of Chinese enterprises. On the other hand, this model is important to protect the interests of investors and creditors. After the financial crisis of2008and the financial regulatory policy to realizing that establishing one timely and effective financial crisis warning system. It is use for to promote the regulatory authorities, listed companies, financial organize gain more opportunity, it is use for to protect creditors.In current, there are two methods of the financial crisis warning, they are traditional method and artificial intelligent method. Traditional statistical method has the advantage of simple, intuitive, analytical expressions can be drawn specifically. It’s easy for readers to understand the meaning of model. The shortcoming of traditional statistical methods is the limitations to variables are very strict, it requires series variables to meet the statistical characteristics, such as the independent variables to normal distribution, and the variables must be independent, with equal variance-covariance matrix. But these assumptions are only a special case of the actual situations. The traditional statistical model can not fit the actual data. The application of artificial neural network is representative to other artificial intelligence method. Although, there are inherent flaws, such as, the whole process is abstract, no specific analytical expression. But this model compensates the shortcomings of traditional statistical method, which are widely used in various fields of study. The artificial neural network is a kind of information processing system. It has been developed form human brain and structure. It uses physical technology to realize the information transmission mechanism of biological neurons.The advantage of artificial neural network in the financial crisis warning is that the whole process is totally objective. In addition, the neural network model is connected with a large number of processing units composed of a wide range of artificial neural networks, and information distributed storage and parallel co-processing, fault tolerant, especially taking into account the need for handling a variety of factors and conditions of the imprecise and fuzzy information processing, pattern recognition and nonlinear mapping problems. It is effective to use the artificial neural network to create the financial crisis warning model. On the basis of reading literature, I compared the relevant literature to fund the advantages and disadvantage of different search methods. By comparison there are still many shortcomings to apply the statistical methods to establish the financial crisis warning system. First, in combination with previous research results and the theory of analysis financial statements analysis to gain the financial indicators of this thesis. At the same time, I define the choose principles of research sample and financial data. I selected100listed companies of the Shanghai stock exchange and the Shenzhen Stock Exchange as the sample of financial distress company. The sample companies had been the first time special treatment from2007to2009, and the features of these companies in according with the definition of the financial distress. At the Sam time, I selected another100companies randomly as the sample of non-financial distress company. I designed this experiment with the financial data belong to the time period of t-2. Second, in order to find the useful financial indicators I had chosen T test method to select the initial indicators. These useful financial indicators are the indicators that have significant differences between the financial distress company and the non-financial distress company. Third, dividing the paired100group companies in to80groups and20groups. The80groups used as the training sample and the20groups used as the test sample. And the useful financial indicators used as the input nodes of the BP neural network financial distress warning system. Then, testing the effectiveness of this system. At the end of this thesis, making the validity conclusion about the BP neural network financial distress warning system. Finally, the neural network financial distress warning system overcomes the limitations of the traditional statistical methods, such as the assumptions and the sample demanding strict the assumptions and sample demanding strict limitations.In the final stage of the research design, the result of test achieves80%accuracy. Based on the accuracy and other relevant factors, obtained in the practical application of early warning model with effectiveness. There are different values to different subjects in the application of the model. The security regulators can use this model to detect the stock companies to maintain the stability of capital markets and investor interests. Banks and non-banking financial institutions use this model to rate the stock companies financial status. The stock company can apply this model to evaluate the level of management. Consulting firms and individual investors can use this model to avoid unnecessary losses.The innovations of this paper:First, novel research perspective; second, the choice of distress samples more reasonable; third, the determination of the output was severely restricted; finally, application of the model proposed.There are some shortcomings for this research worthy to improve. So this research proposes some suggestions for further research:(1) the definition of the financial crisis can be further explored.(2) The sample selection can be limited to a specific industry.(3)The structure of the neural network is worth exploring.(4) The determination of the output node is also worth exploring.