Financial Fraud Detection By Using Machine Learning

To be with a listed company that has good performance and sound financial affairs or a public company which contents financial fraud statements,that is the primary question for the investors all over the world.The listed companies which malevolently created financial fraud statements in the past eleven years since the 2008 financial crisis have brought tremendous harm to the capital market investors of the entire world.In sharp contrast to the capital market participants' horrible lack of ability in financial fraud statement identification,the listed companies are constantly and relentlessly cheating their shareholders.This makes it necessary to create an efficient,accurate and simple method to identify the financial fraud statements of listed companies.Based on the financial data of 137 listed companies from 2013 to 2018,this paper makes an empirical study on the financial statements falsification of listed companies using Machine Learning.Through the empirical analysis such as Logistic Regression,Gradient Boosting Decision Tree and Random Forest,the primary testing results are concluded,then using Random Forest model for the secondary testing,in this paper,an efficient,accurate and simple comprehensive application model of machine learning is constructed.The empirical results show that the net interest rate of sales is a good indicator to identify the financial fraud statements of listed companies.The operating profit margin,the earnings per share and the net profit margin of the financial fraud listed companies are abnormal especially in the year that the financial statements fraud happened or one to two years before.The Machine Learning model such as Logistic Regression is not effective in detecting the financial statements fraud of listed companies.The comprehensive Machine Learning model constructed in this paper achieves the accuracy of 97.14% in judging the financial statements of listed companies.It provides an accurate and practical method for identifying the financial statements of listed companies.It is of great practical significance for investors and securities research institutions in dealing with financial fraud.