Study On Method Of Listed Companies’ Financial Distress Prediction Based On Dempster-Shafer Evidence Theory

Financial distress will bring enterprises, investment institutions, social andgovernment great cost. Therefore, it is important to predict financial distress accuratelyand timely, which can help enterprises to prevent or dissolve the financial crisis, help thefinancial investment institutions to avoid investment risk and help government regulatorsto regulate and maintain the market more scientifically. Financial distress prediction,which became popular in foreign countries in the early1960s, is a classical topic.However, the backwardness of the domestic modern corporate and finance systems leadfinancial distress prediction for a late start in the domestic. Thus, it is still of greatimportance to search for better methods of financial distress prediction which are suitablefor domestic listed companies.Under this background, this paper extends the research of financial distressprediction so as to improve the prediction accuracy of listed companies’ financial distress.Firstly, exciting literatures have no consistent conclusion on the view whether it canimprove the prediction accuracy by reducing the financial ratios in the area of financialdistress prediction research. Based on related literatures, this paper uses Dempster-Shaferevidence theory to fuse the outputs of independent prediction in decision layer. Therefore,it not only retains the information of all financial indexes, but also it avoids the computingcomplexity of prediction models. We verify the validity of proposed method in this paperthrough the sample of Chinese listed companies. The empirical results show that proposedmethod can greatly improve the prediction accuracy of companies in financial distresswithout lowering the average prediction accuracy.Secondly, combination prediction methods have higher prediction accuracy andstability than single prediction methods. Based on some exiting combination predictionmethods, this paper proposes a new combination prediction method by usingDempster-Shafer evidence theory and rough set. Rough set automatically determines theweight of each single prediction method according to the outputs of single predictionmethod. And we use Dempster-Shafer evidence theory to combine the outputs in decisionlayer. We verify the performance of proposed method using the sample of Chinese listedcompanies. By comparing with some single prediction methods and other combinationprediction methods, the empirical results show that proposed method not only has higherprediction accuracy for companies in financial distress, but also it has greatly improved the prediction accuracy for total sample including health companies and companies infinancial distress.