财务报告舞弊识别的实证研究

随着舞弊手段日趋隐蔽，需要建立一个快速有效的识别模型。本文选取了24个相关指标作为解释变量，选取了72家财务报告舞弊上市公司作为样本。首先建立Logistic回归模型，模型的判断率达到了71．7％。然后将Logistie回归得到的7个解释变量，用运到BP神经网络模型之中，使判则率达到87．1％，高于出了16％。说明13I）神经网络的总体识别率更高。

The Empirical Study of Financial Reporting Fraud

With the increasing concealment for fraud ways, a rapid and efficient recognition model need to be established. In this paper,24 relative indicators are selected as explanatory variables;72 China＇s list compa- nies which had financial reporting fraud are selected as samples. Firstly, a logistic regression model is built, and the recognition rate of this model is 71.7 %. Seven explanatory variables from logistic regression is applied to BP network model. The recognition rae get to 87.1% ,which is 16% higher than the logistic regression model. It indicates that the general recognition rate of BP neural network model is better.