Financial Early-warning Model For Manufactural Listing Companies By Using Adaboost

With the rapid development of China's securities market for more than ten years, it has provided listing and financing opportunities for nearly 2000 enterprises. However, these companies meet not only the development for golden future, but also fierce competition.Some of them were even been special treated by the commission. On one hand, this will affect the existence of these companies, on the other hand, it will also bring enormous lose to related benefits. Therefore , no matter to the companies themselves or to the investors, creditors, regulators, an effective forecasting model for finance crisis is needed. It could help them prepared in advance to tackle the crisis.Domestic researchers has already established finance crisis forecasting model by using artificial neural network, and have achieved good results. However, single neural network's prediction is some sort of precarious, this paper is going to use adaboost algorithm(which can improve any weak learners to strong ones)combine with neural networks to establish early-warning model. The research process of this thesis is as follows: first, on the basis of related thesis about finance crisis prediction, we choose sixteen variables for research;35 manufacturing listed companies which were special treated for the first time because of finance exception from 2007 to 2009,and their corresponding normal manufacturing listed companies are selected as comparative samples, based on financial data ahead of 2-3 year of each company is special treated. Then, principal component analysis is used to generate 4 principal components as inputs of the forewarning model. Finally, we use adaboost algorithm and matlab programming establishing the required model to predict crisis, with BP neural network as weak classifier.From the research result, we find out that this BP-Adaboost model's forewarning accuracy is much higher than the average prediction accuracy of six BP neural networks. To those companies which are alerted, they can improve their management from the previously related four aspects. In a word, adaboost algorithm is feasible applied to financial crisis forewarning, it can help us building a better warning model.