本研究應用資料探勘中貝氏信度網路、支援向量機以及決策樹等方法，並採用財務及非財務變數，作為協助偵測財務報表舞弊之工具。研究對象為1998年～2005年60家發生財務報表舞弊及非財務報表舞弊之公司。結果發現，財務及非財務資訊有效用於辨別財務報表舞弊；且貝氏信度網路分類效果最好，次為支援向量機，最後為決策樹。
This paper explores the effectiveness of Data Mining Classification techniques such as Bayesian Belief Networks, Decision Tree and Support Vector Machine in detecting firms that issue fraudulent financial statements (FFS) and deals with the identification of factors associated with FFS. First, we underline the importance of financial and non-financial factors that can be used in the identification of FFS. Second, a number of experiments have been conducted using these techniques which were optimized using a data set of 60 fraud and non-fraud firms in the recent period 1998~2005. The results shows that the Bayesian Belief Network has better performance than the Decision Tree and Support Vector Machine.
關聯:
Asian Journal of Management and Humanity Sciences 3卷1-4期 P.15-30
