The Study On Early-warning Of Financial Crisis For Listed Manufacturing Companies Based On BP Neural Network

With development of the market economy and expansion of the stock market, the competitions among enterprises are becoming drasticer. The enterprises are always facing the crisis to risk business failure or declare bankruptcy. To listed companies, if falling into the financial crisis, it will bring enormous loss to not only itself but also investors and creditors. So it has very realistic and influencing meanings to establish an effective practical early-warning model.The thesis constructs a financial early-warning research system for listed companies based on BP neural network descriminant technique. 150 manufacturing companies of same development stage, same size, also same time window are being selected as research objects, and being used in the empirical process.The features of the study are as follows:Firstly, unlike previous study, the paper uses index system which has been screened twice from 25 complete financial indexes instead of some indexs set chose randomly or empirically. The screening process makes the forecast more reasonable without decreasing the amount of information, also makes the model more easily calculated and practical.Secondly, respectively use three methods which are traditional statistical discrimination model, artificial neural network model and the combination discrimination model to test on financial early-warning, make comparison and analysis. The result shows BP neural network prediction method makes full use of fault tolerance and self-learning nature of artifical neural network with high accuracy and applicability. The combination forecast model inherits the advantages both of statistical method and BP neural network. It has good prediction accuracy as well while no obivious strength comparing to the pure neural network model with more complexity.