摘要
以2005年A股制造业的36个财务欺诈公司及相应的配对公司为样本,采用23个财务指标,通过数据挖掘途径,运用多元判别分析、支持向量机、决策树和自己设计的集成分类方法构建了财务欺诈识别模型。实证结果表明:4种模型都具有一定有效性,集成分类方法的识别准确度最高且识别效果最好。
Based on 36 A-shared manufacturing listed companies with fraud financial report and the same number of matched sample with normal financial report in 2005 ,financial fraut identification models are built using 23 financial indicators and four including multiple discriminative analysis,support vector machine and an integrated classifier.The result indicates that the four models are all effective to some extent and the integrated classifier outperforms other methods in terms of prediction accuracy.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第34期226-230,共5页
Computer Engineering and Applications
基金
国家教育部新世纪人才支持计划No.NCET-04-415
教育部哲学社会科学研究重大课题攻关项目(No.07JZD0020)
上海市教育委员会科研创新项目(No.08ZS33)~~
关键词
财务欺诈识别
支持向量机
决策树
集成分类方法
financial fraud identification
support vector machine
decision tree
integrated classifier