财务报表窗饰程度之侦测模型.pdf

財務報表窗飾程度之偵測模型 *林欣瑾 中國 文化大學會計學系 * Annman1204@ 徐銘甫 國立暨南國際大學 資訊管理學系 Hsumf0222@ 張清和 國立暨南國際大學 經濟學系 chc6175@ 摘要： 近年來許多上市櫃公司接連發生會計醜聞，其 大多採用盈餘管理的技術美化 財務報表，使投資大眾認為公司的營運沒有異常 並降低其的疑慮，由此可見， 盈 餘管理的偵測是有其困難度的，本研究將採用一種新穎的偵測模型，此模型採用 模糊理論、約略集合以及蟻群演算法去建構出一個預測模型，模糊理論可以用來 克服約略集合於離散化資料時所引起的資訊遺漏的問題，約略集合可以用來建構 預測模式，但找尋約略集合最佳的子集是相當耗時的，因此，本研究將其轉化為 最佳化的問題並採用蟻群演算法去處理，實證結果指出，本研究所提出的模型具 有相當優越的預測效力，可以提供投資大眾另一個決策輔助的工具。 關鍵字 : 盈餘管理，偵測模型，決策分析 A model for earnings management detection Abstract: Numerous financial failure cases related to the listed companies have increased significantly, but related parties and investors are complicated to predict the financial failure, especially in the earnings management. Earning management is manipulating earnings to reach the management’s purposes implementing some avenues or methods. Most proportion of the prior studies merely focused on determining the related elements which can affect earning management significantly. A few studies laid emphasis on earning management forecasting model construction. Thus, this study proposed an emerging detecting model use to forecast the level of earning management. The emerging model consists of fuzzy set theory, rough set theory and ant colony optimization. The fuzzy set theory was used to overcome the problem of information loss. The rough set theory was implemented to construct the forecasting model. The inherent drawback of rough set theory was difficult to determine the best reduct. Thus, the best reduct determination was transformed into optimizat