摘要
本文从CCER数据库中的上市公司违法违规数据库取得2006-2009年利润操纵上市公司样本,根据上市公司关键财务数据,构建配对样品组。文章运用粗糙集简化利润操纵的识别指标,通过蒙特卡洛模拟,提出基于BP神经网络模型的中国上市公司利润操纵的识别模型。本研究还引入DEA效率指标,建立改进的中国上市公司利润操纵识别模型,有效降低模型的第二类错误,从而将模型的正判率从71.43%提高到85.71%。
In this paper, listed companies with profit manipulation are collected from the illegal database in CCER database in 2006-2009. Then matched sample groups are established based on the key financial data. By using the rough set theory, indicators of profit manipulation are simplified. We establish a BP neural network model by using Matlab program for the first time. Besides, this paper also uses DEA efficient index to establish an improved model of Chinese listed companies to recognize profit manipulation. The second type of error of the model is reduced and the correct rate of the model is successfully improved from 71.43% to 85.71%.
出处
《数理统计与管理》
CSSCI
北大核心
2013年第3期440-451,共12页
Journal of Applied Statistics and Management
基金
2012年度中国社会科学院青年科研启动基金项目的资助