Model Design In Bankrupt Prediction Based On Neural Networks

Financial distress is a world problem. Because it greatly affects the financial decision making of investors, credits, and bank officers. Auditors also need information on financial distress prediction for the goning-concern judgment. It is a gradual process for a company getting into financial distress, so it must have some signs and thus could be forecasted. Forecasting corporate financial distress accurately makes much sense. For example, the exact forecast could increase interests of investors and debtors, could protect managers from financial crisis, and could help government to monitor listed company's quality and securities market's risk. Since 1960s, studies on corporate failure prediction have prebailed both in the USA and in European countries. Financial distress predicting has played more important role in securities investment, credit risk mangment, auditors'decision making and corporate financial mangment overseas than everbefore.There are many statistical procedures to handle this financial distress prediction problem. The most widely used classification technique is statistical methods including MDA, logit, and Probit methods. MDA has been widely applied to the business classification, including bankruptcy prediction, credit rating, and bank loan classification, etc. But the violation of the underlying normality assumption of independent variables causes the biased results. As one of alternative methods, neural network approach represents a nonlinear discriminant function as a pattern of connections between its processing units, so it is very promising for the financial distress prediction problem.few researches on financial distress predictions have been conducted in our country until very recently. Using the characteristic of a higher order derivative of some base functions can be expressed by their lower order derivatives, a cascade-correlation algorithm with tunable activation functions is proposed in this paper. And variables are selscted by Dynamic principal component analysis in this financial distress prediction. This paper make use of the financial statements of listed companies in China Stock Exchange(1998-2002) and build a new model for corporate failure discrimation by ensemble method(BP,CC,TAFCC)and dynamic principal component analysis. The results show that ensemble method prediction ability is superior than MDA, logit ect.