Lib-SVMs Detection Model of Regulating-Profits Financial Statement Fraud Using Data of Chinese Listed Companies
LI Xiu-zhi, Shuangshuang Ying · 2010
This paper uses Lib-SVM algorithm of RBF kernel and linear kernel to develop a model for detecting regulating-profits financial statement fraud with the data of 112 Chinese listed companies. It turns out that the prediction accuracy of Lib-SVM algorithm for RBF kernel function model is 86.667%, the overall accuracy is 87.5%. And the prediction accuracy of the Lib-SVM linear kernel function model is 83.333%, the overall accuracy rate is 86.612%. With the same samples, a Logistic regression model is developed, and the corresponding accuracy is 80% and 83.036%. The study reinforces the validity and efficiency of Lib-SVM algorithm as a research tool and provides additional empirical evidence regarding the merits of suggested variables for regulating-profits fraudulent financial statements.