摘要
随着信息技术的不断发展和财务数据的急剧增加,财务报表舞弊手段愈发高明和隐蔽,已有基于统计的检测方法已不能对财务报表舞弊进行有效检测,本文讨论在财务指标区间离散化的基础上,运用关联分析技术挖掘财务指标与财务舞弊之间的隐含关联关系。首先,介绍财务报表样本和特征指标选择准则,从权威数据库中分别选取了234份财务舞弊样本和234份控制样本,得到13个初始变量。然后,采用基于单纯的拐点区间划分和基于聚类的拐点区间划分方法对样本进行属性离散化,可分别得到59和82个离散变量。最后,运用频繁项集挖掘算法挖掘出财务指标与财务舞弊之间的关联规则,并用于对样本进行识别。实验结果表明,利用基于聚类技术的拐点区间离散化挖掘得到的关联规则可提高模型的识别准确率。
With the continuous development of the information technology and the dramatic increase of finan- cial data, the ways of financial statement fraud become more advanced and hidden than ever, and the existing statistics-based detection methods cannot detect the fraud effectively. Based on the discussion on discretization of financial indicators, the paper uses correlation analysis technology to mine the implicit relationships between the financial indicators and financial fraud. First, the paper imroduces the selection criteria of financial statement sam- pies and feature indices, and gets 13 initial variables from 234 financial fraud samples and 234 control samples se- lected from the authoritative databases. Then, 59 and 82 discrete variables are gained through attribute discretiza- tion using the simple inflection point interval partition technology and the inflection point interval partition method based on clustering respectively. Finally, the algorithm of frequent item set mining is used to mine the association rules between the financial indicators and financial fraud. Experimental results show that the detection model using the inflection point interval partition method based on clustering can improve detection accuracy.
出处
《科技广场》
2013年第2期240-244,共5页
Science Mosaic
基金
大学生创新创业训练计划项目资助(编号:201210421031)
关键词
财务舞弊
关联规则
财务报表
拐点区间划分
Financial Fraud
Association Rules
Financial Statements
Inflection Point Interval Partition
