Research On Financial Fraud Recognition Of Listed Companies Based On Data Mining

Chinese market economy is in good development, along with which the number of listed companies keeps rising. Chinese economic is in the stage of rapid development and the system is not perfect, so the status of listed company financial fraud is more and more serious. Financial fraud incidents occur more and more. While they are not prohibited, even though Chinese government issued a series of regulations endeavoring to cut down financial fraud phenomenon of listed companies. Financial fraud has great harm, since it can not only make the investors suffer from wrong investment decisions on the basis of financial disclosure of corporate operation status, but also do harm to the securities market and the credit of the whole nation. So, the study of the listed company financial fraud identification has very important practical significance.For the governance of financial fraud, people try to look up appropriate and reasonable method.Domestic and aboard scholars no longer focus on normal research of financial fraud, they tend to concern empirical research which can provide more help to people. Based on the database of listed corporation, this paper established an indices system of listing corporation financial fraud identification through index evaluation, and then established the listing corporation financial fraud identification model by employing an ensemble method based on three data mining algorithms,resorting to the data mining tool of IBM SPSS Modeler 14.01. First of all, a sample set was extracted from the Resset database including 74 listed corporation as financial fraud companies because of administrative penalty and 74 counterparts as non-fraud companies. Then, 16 indicators have a significant role for financial fraud identification were screened from 28 Financial and non-financial alternative indicators. And then, 5 principal factors extracted from the 16 indicators through factor analysis as the input variables, and fraud or not(marked as 0, 1 respectively) as the output variables, a financial fraud identification model was constructed based on an ensemble method integrating three algorithms including neural network, decision tree and support vector machine, which was tested afterwards. Test results show that this model can effectively identify the listed corporation financial fraud samples, which can be applied to financial fraud identification of listed corporations and can be a deterrent to fraud motivation of listed corporations.