基于非线性-主成分Logistic回归的会计舞弊识别研究

混沌理论认为,人类行为大多具有非线性特征。会计舞弊属于行为会计的研究范畴,而传统上基于统计理论构建的舞弊识别模型大多受限于线性约束假设,可能存在模型设定偏误和信息提取不充分的缺陷。以沪深A股受到监管处罚的上市公司及其配对公司为样本,借鉴Taylor展开式的非线性思想,并使用主成分分析消除变量多重共线性,构建了非线性-主成分Logistic回归的会计舞弊识别模型。与线性回归模型对比发现,前者具有更高的舞弊识别正确率,模型拟合度更优。应用这一模型有助于更加充分提取舞弊识别信息,提高舞弊识别效率。

Using Nonlinear-Principal Component Logistic Regression for Accounting Fraud Identification

Chaos theory suggested that most of human behavior appeared to be non‐linear .Accounting fraud belonged to the field of behavior accounting . Traditionally ,fraud identification model based on statistical theory limited to linear constraint assumptions mostly . T here may be such defects as model specification errors and information extraction insufficiently . It chose Shanghai and Shenzhen A‐share listed companies subject to regulatory sanctions and matching companies as samples . Based on the nonlinear ideology of Taylor expansion , and the principal component analysis to eliminate variables multicollinearity ,it constructed a nonlinear‐principal component Logistic regression of accounting fraud recognition model . The model has a higher recognition accuracy ratio , more reliability on parameter estimation and has higher goodness of fit than the linear regression model .The model is helpful to extract fraud identification information more fully ,and improves the efficiency of fraud identification .