稀疏组Lasso-logistic回归模型在财务报告舞弊识别中的应用研究  

Research on Identification of Fraud in Financial Reporting Based on Sparse Group Lasso-logistic Regression Model

作　　者：王威 WANG Wei(Business School,Guilin Tourism University,GuiLin 541006,China)

机构地区：[1]桂林旅游学院商学院,广西桂林541006

出　　处：《数学的实践与认识》2020年第9期49-58,共10页Mathematics in Practice and Theory

基　　金：桂林旅游学院科研启动基金(2018QD007);广西高校中青年教师科研基础能力提升项目(2020ky22010)。

摘　　要：财务报告舞弊行为对广大投资者的切身利益造成巨大损害,如何高效识别财务报告中的舞弊行为已成为目前研究的热点.在对已有的财务报告舞弊识别模型分析的基础上,提出一种基于稀疏组Lasso-logistic回归的识别模型,并通过选取近8年间180家上市公司年报数据作为样本,综合财务及非财务指标,从盈利能力、营运能力、偿债能力、治理结构等方面设计了15组29个解释变量使用该模型进行了实证研究.结果证明,与以往的向前Logistic回归、Lasso-logistic回归和组Lasso-logistic回归等模型相比,稀疏组Lasso-logistic回归识别模型不但具备良好的变量筛选能力而且可以获得更好的识别效果,具有较高的应用价值.Fraud in financial reporting had caused great damage to the vital interests of investors.How to effectively identify fraud in financial reporting had become a hot research topic.Based on the analysis of the existing financial report fraud identification model,this paper proposes a sparse group Lasso-logistic regression identification model.By selecting180 listed companies’ annual report data in the past 7 years as samples and integrating financial and non-financial indicators,15 groups of 29 identification variables are designed from profitability,operating ability,solvency,governance structure and other aspects to conduct empirical research using this model.The results show that,compared with previous models such as forward Logistic regression,Lasso-logistic regression and group Lasso-logistic regression,sparse group Lasso-logistic regression identification model not only had good variable screening ability but also can obtain better identification effect,and had higher application value.

关 键 词：财务报告 舞弊识别 稀疏组Lasso LOGISTIC回归 

分 类 号：F275[经济管理—企业管理] F832.51[经济管理—国民经济] F224