An Effective Financial Statements Fraud Detection Model

Abstract

The purpose of this study is to establish an effective and rigorous financial statements fraud (FSF) detection model. The samples of this study are the listed companies in Taiwan, totaling 220 companies and including 55 companies with financial statements fraud and 165 normal companies. The data are from the Taiwan Economic Journal (TEJ) during the period from 2006 to 2015. In Stage I, decision tree CART and artificial neural network (ANN) are applied to select the important variables. In Stage II, decision tree CHAID, ANN, and support vector machine (SVM) are used for modeling. The results show that the ANN-CHAID model has the highest FSF detection accuracy of 87.41%.