Empirical Research Of Financial Distress Prediction Model In Manufacturing Industry In China

With the reform of China’s economic system, China’s securities market has had aflourishing development. In recent years, the number of listed companies has had arapid growth, but China’s securities market is still a lot of risk that cannot be ignored.The international debt crisis in Europe and the lower of credit rating in U.S. all have agreat impact on China’s securities market. To the beginning of this year, China’s Ashares are not yet getting out of the haze and the number of listed companies aregetting into capital crisis situation. The complex situation and the listed companycompetition have exacerbated the risk of listed company. Therefore, the establishmentof an effective financial distress warning model has an important significance. On theone hand, it plays a role to improve the ability of corporate managers to avoid trouble;on the other hand, it is also conductive to creditors, investors and stakeholders tomake good choices according to r the early warning results.The purpose of this study is to find a more accurate model in order to forecastthe financial situation of listed companies. The structure and the specific contents ofthis article can be shown in the following aspects.In Chapter1, the article briefly introduces the research background and researchsignificance. And then it gives a brief introduction of financial plight definition.Lastly, it reviews the current situation of financial distress warning model.In Chapter2, this article introduces several models, such as Fisherdiscriminant model, Logistic regression model, BP neural network model, SVMmodel, and weighted voting combination model. In addition, the article makes asimple comparison between the screening methods of indicators. In the context, theauthor introduces the theory of the neighborhood rough set.In Chapter3, first complete the construction and selection of samples and indexsystem for this study, and second process the data for build the data into the model. The samples in this article are selected from the A-share listed companies.358companies initially are collected by author. And ultimately346companies are rest bydeleting some companies (of which77are ST companies, and269are normalcompanies). The ST’s year of company is T and in this case this article is to study theT-3years’ data. There are30primary indicators in this article. It not only includesfinancial indicators, but also contains corporate governance indicators, stock agencies,market income and company size. Then process the data in theory of theneighborhood rough set.In Chapter4, the author mainly introduces the empirical analysis. The first stepthe author use the Fisher discriminant model, Logistic regression model, BP neuralnetwork model and SVM model to establish a single financial distress predictionmodel. In this context, the author builds a combined model of weighted voting.Finally， research and compare empirical results empirical results based a singlemodel and combined model.In this paper, the research conclusions can be easily concluded in the followingaspects. First, in two statistical modeling methods, the total asset turnover has a bigeffect on the financial distress warning model. At the same time, the asset turnoverreflects the company’s operations. Therefore, a company operational capability has avery important influence on company development. Second, for a single classifier, theprediction ability of BP neural network is higher than Fisher discriminantanalysis and Logistic regression models. In artificial intelligence methods, theprediction ability of SVM model is higher than BP neural network; in statisticalmethods, the prediction ability of Logistic regression model is higher than Fisherdiscriminant analysis model. Third, the prediction ability of financial plight weightedvoting combination forecasting model is also higher than the support vector model.So it has a clear superiority in Financial Distress Prediction.