Research On Financial Distress Prediction Of Listed Companies Based On Neighborhood Rough Sets-Neural Network

With the advent of the information age, enterprises are facing greater uncertainty, and the risks. In the meantime,because of abnormal financial situation, the number of listed companies being special treated has increased dramatically. Early-warning model of financial distress, as a forecasting model, can provide reference for companies strategic adjustments, investors investment decisions, and creditors debt issuance. Therefore, it is of great theoretical and practical significance that to carry on this research and build an efficient model.Based on the definition of financial distress, the analysis and the appraisal to the classical document of the domestic and foreign, this thesis pointed out the main inadequacy of the existing studies. Deeply analysing the current situation of the financial crisis that listed companies met and summarizing the causes of financial disress, exploring the relationship between corporate governance and financial distress, moreover the financial and non-financial distress characteristics were researched. All of that supply theory and realism basic for the index choice of financial distress forecasting model. After that, The application of the principles and knowledge reduction algorithm of Neighborhood Rough Set would be discussed, and at the same time, the theory of neutral network will be introduced in detail. Based on above, we posed a further analysis of the basic principles of the combination, and built a hybrid model that combined Neighborhood Rough Set and Neural Network. Firstly, this paper dealt with the data primarily to simplify index system by neighborhood rough set as neutral network. And then setting the simplified index system as the input of neutral network to make categorical forecasting. In the selected aspects of forecasting indicators, based on financial factors, the researcher add some non-financial indicators into the model, which can reflect the corporate governance and audit opinions to enable the predictors can be more comprehensive and better to reflect the company's business situation.At last, In order to avoid the impacts of industry characteristics, this paper choose data from A share listed company annual reports from 2003 to 2009, and conduct an empirical research on the hybrid pre-warning model. Empirical results show that the model can effectively remove redundant information, and compared to the single neural network model and Logistic regression model, its forecast accuracy has increased with shorten training time and higher efficiency.