Study On Enterprise Financial Distress Prediction Model Based On Optimization BP Neural Network

China’s accession to the WTO, is also facing unprecedented challenges in the fierce market competition, enterprises are facing a wave of global economic integration today. At any time are faced with the choice of the failure and plunged into the predicament and even the point of bankruptcy. China’s listed companies, if caught in a dilemma, not just to bring the plight of their own development, will bring economic losses to investors and creditors, will seriously affect the country’s economic production and living. So China’s listed companies, financial early warning is necessary.In this thesis, first a brief introduction to the financial condition of listed companies in China, and through the collation of the relevant literature, global scholars Financial Distress of study a simple comb, a review of the past century research results; analyze and compare the final choice to the highest predictive accuracy of BP neural network as a financial early warning model, elaborate and the operating principle of the model and the reasoning process and analyze its shortcomings and propose solutions. Then optimize BP neural network training learning proved by the experimental results of the test software, lack of BP neural network is a good solution.Finally2007-2011listed company financial data early warning analysis study, we found that, created by BP neural artificial network prediction model, of the selected1207inside the normal class of the company’s financial predicament predict the results of accurate rate more than80%, thus proving the validity of the BP artificial neural network model. Accounted for the vast majority of the proportion of the number of normal class company model samples, sample number of serious imbalance caused by too much "memory" normal class company BP upgrade network training, resulting tend to first loss year and ST mistaken for normal class company. And the accuracy of the predictive ability of the BP neural network model built on the company’s first loss class is the next research focus.