| Securities market is becoming more and more prosperous and successful in China,meanwhile we also realize the importance of accounting information.There are many problems in this market,especially in the accounting fraud field.The market cannot play the resource allocation function,so it is not be trusted by stakeholders.Accounting fraud becomes more subtle,so we cannot find the problem in the financial reporting.It is important for stakeholders to find the accounting fraud on data mining technology.We put forward the accounting fraud identification indexes in six aspects on the basis of the accounting fraud definition and methods.This paper selects the data mining of the classification,after analyzing the present data mining technology.This paper clarifies how to choose the sample and variable on the basis of the characteristics of accounting information.After data preparation work,we select 19 financial indicators and 138 samples to construct accounting fraud identification.This paper select Logistic regression,decision tree and support vector machine(SVM)in the accounting fraud identification,which discusses the recognition effect,extrapolation effect and prediction effect of the accounting fraud model.This paper also compares and analyzes the variables and recognition effect of the accounting fraud model.Studies have shown that Logistic regression,decision tree and SVM have good recognition effect in the accounting fraud.The decision tree recognition effect is best.Recognition accuracy of these model have a certain degree of decline in the prediction effect test.The decision tree prediction effect is better,which is higher than the other two ways.The paper gets the three main conclusions.Is closely related to the accounting fraud in the financial report.Financial indicators of listed companies have strong ability of accounting fraud identification.Logistic regression,decision tree and SVM have good prediction effect of accounting fraud. |
PDF Full Text Request