基于分类回归树算法的上市公司会计信息失真识别研究

利用26个财务变量建立分类回归树模型对会计信息失真进行识别研究,结果表明所建模型对会计信息失真企业的正确识别率达到80%以上,能将第二类错误率控制在20%以下.实证还发现留存收益在总资产中的比率小于2%的公司很容易出现会计信息失真,最后作者利用8年数据对该结果进行检验,表明其识别能力非常出色.

Identification of False Financial Statements by Classification and Regression Trees

In this paper,we established a CART model with 26 financial variables,and found that: the CART model can achieve over 80% correct recognition rate for distortion accounting information and control the second class mistake under 20%.We also found that companies whose ratio of retained earnings to total assets is less than 2% are easily involved in delivering a fraudulent financial statement to the public,and this result is proved to be effective with data from 2000 to 2007.