Research On Multi-Category Enterprise Financial Distress Prediction Based On Decomposition And Fusion Strategy And Support Vector Machine

When the internal management is chaotic or suffers serious external shocks,the company's financial situation is in a dangerous situation,which will affect the company's sustainable operation.The financial distress of a company is a comprehensive manifestation of the deterioration of various internal and external risk factors.Even the crisis caused by non-financial factors will eventually be presented in the form of financial distress.Building an early warning system that can effectively predict the financial situation is an important part of enterprise risk management.The multi-category enterprise financial distress prediction model studied in this paper will be based on decomposition and fusion strategy and support vector machine.This paper breaks through the thinking that the financial status of enterprises can only be divided into "good finance" and "financial distress".The enterprises with good finance can be classified into" Financial Sound Enterprises "and "Financial Pseudo-good Enterprises ".The enterprises with financial distress can be classified into "Financial Moderate-crisis Enterprises" and "Financial Serious-crisis Enterprises".That is to say,the financial status of enterprises can be divided into four categories.With the increase of enterprise financial status categories,the training of multi-category enterprise financial distress prediction model becomes more complex.Traditional Support Vector Machine(SVM)methods can't be directly applied to the training of multi-category financial distress prediction models.Therefore,this paper combines decomposition and fusion strategy with basic SVM algorithm to build a multi-category enterprise financial distress prediction model.The decomposition and fusion strategy is One-Versus-One(OVO),One-Versus-Rest(OVR)and Error Correcting Output Coding(ECOC).Three kinds of multi-category enterprise FPD models,OVO-SVMs,OVR-SVMs and ECOC-SVMs,are constructed.The financial data of A-share listed companies in Shanghai and Shenzhen Stock Exchange are used for empirical research.The results show that the average prediction accuracy of the three models is higher than 80%,in terms of the overall and four types of financial status.The three kinds of multi-category enterprise financial distress prediction models constructed in this paper have their own advantages and disadvantages.Users can choose appropriate models according to specific needs.The financial index system used to construct the model has certain identifiability and can be used for auxiliary monitoring.At the same time,it shows that there are many kinds of financial status of enterprises,so we should be alert to the multi-types of financial status of enterprises.Financial Sound Enterprises are the safer ones with sustainable development,while the Financial Pseudo-good Enterprises,Financial Moderate-crisis Enterprises and Financial Serious-crisis Enterprises belong to the enterprises with different degrees of financial distress,which need to be vigilant.The main innovations of this paper are as follows:?Breaking through the restrictions of dividing enterprises into "good financial situation" and "financial distress",this paper divides the financial status of enterprises into four types.?Creatively build multi-category enterprise financial distress prediction model based on decomposition and fusion strategy and support vector organization.The multi-category financial distress prediction model constructed in this paper is more close to the actual situation of Chinese enterprises.The results of this study not only help to enrich the theoretical system of FDP methods,but also provide important decision support tools for business managers,investors and creditors.