基于Adaptive Lasso-Logistic回归模型的财务报告舞弊识别研究

A Study of Identification of Fraud in Financial Reporting Based on Adaptive Lasso-logistic Regression Model

财务报告舞弊行为会对广大投资者的切身利益造成巨大损害,因而如何高效识别财务报告中的舞弊行为已成为目前学界研究的热点。本文在对已有财务报告舞弊识别模型进行分析的基础上,针对Logistic回归模型在财务报告舞弊识别中存在的变量多重共线性和计算复杂等问题,提出Adaptive Lasso-Logistic回归识别模型,并以2010年—2017年间我国320家上市公司的年度财务报表数据为样本,从盈利能力、营运能力、偿债能力等方面设计了18个财务比率指标进行实证研究。结果表明,与传统的Logistic回归模型和Lasso-Logistic回归模型相比,Adaptive Lasso-Logistic回归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。

The fraud of financial report has caused great damage to the interests of investors.How to identify the fraud in financial report efficiently has become a hot research topic.Aiming at the problems of multicollinearity of variables and computational complexity in logistic regression model of financial report fraud identification at present,this paper established an identification model based on Adaptive Lasso-Logistic regression and applied it to the identification of financial reporting fraud of listed companies.It selected the annual report data of 320 listed companies from 2010 to 2017 as samples,and from the profitability,operating capacity,solvency and other aspects of the design of eighteen financial indicators for empirical research.After comparing the recognition effect with that of full variable regression and Lasso-logistic Regression Model,it is found that the former has not only good variable selection ability,but also the best recognition effect.