A Study On Identifying Fraudulent Financial Reporting Of Listed Companies Based On SVM Model

Enterprises'financial and managerial efficiency are transfered to stakeholders through financial information, which would help them to know enterprises'operating and financial status in the past, present and future. The development of security market is seriously affected by the fraudulent financial information of listed companies in China, which regulation organizations pay attention to, as well as more and more investors. It is useful to develop a model effectively identifying the fraudulent financial information. There are two methods to develop the model: parametric method and non-parametric method. because parametric method is subjected to the population distribution hypothesis, It is unable to study by itsesf, adjust efficiently and correct the model, which make identifying result not satisfactory. The non-linear relationship between fraudulent variables and the fraudulent financial information can been reflected in the non-parametric method, but the non-parametric method has inferior commonality and can't be effectively promoted.Firstly, the background and literature survey of this research is introduced, the purpose, process and structure of this paper is clarified. Then, the theories about fraudulent financial reporting are introduced, which include concept of fraudulent financial reporting, the methods for indentifying fraudulent financial reporting, the variables of identifying fraudulent financial reporting and SVM theory. On the base, the latest brand of the statistical learning theory, support vector machine, is employed to identify the fraudulent financial reporting, 106 listed companies with fraudulent financial reporting and 106 listed companies with true financial reporting is choosed as research samples.Two models indentifying fraudulent financial reporting are developed, one is only based on financial variables, the other one is based on both financial variables and non-financial variables. The author compares their precision with the MDA model. Conclusion summarizes the result, probes into its academic and practical significance, and points out the limitation and direction of the future empirical research.The result shows that the precision of identifying model can be improved by introducing non-financial variables, non-financial variables should be adopt when justifying whether there is fraudulent information in listed companies'financial reports. Meanwhile, the author find that the result of model based on SVM is more precise than MDA model. So the model based on SVM can effectively help regulation organization and investor to identify the fraudulent financial reporting of listed companies.