透過您的圖書館登入
IP:3.141.100.120
  • 學位論文

資料探勘技術建構公司財務預警模式之研究

Data mining techniques build the company's financial crisis early-warning model of research

指導教授 : 黃劭彥
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


本研究運用資料探勘演算法來建構公司財務危機預警模式,選取在臺灣經濟新報資料庫(TEJ)資料。以一家危機公司與三家正常公司1:3的比例,作為配對樣本。樣本變數組合分為「財務指標」、「公司治理指標」以及合併「財務及公司治理指標」等面向,各面向各包含年度別模式:完整前三年、前一年、前二年、前三年以及狀況別模式:實質財務危機、非實質(準)財務危機。再就資料探勘技術演算法建構合計十八個模式的財務危機預警模式。 本研究結論如下:一、倒傳遞類神經網路(Back Propagation Network,BPN) 模式效果最佳,將財務危機之預測能力提昇達九成以上。二、實證「公司治理指標」的確為領先指標,若結合「財務指標」與「公司治理」指標更佳。三、印證單一年度別的精確度比較,除了前三年的「合併財務指標與公司治理指標」略遜0.28%之外,餘皆高於完整前三年,應與財務危機事件發生的時距最大化(Max)所致。四、狀況別模式中印證非實質(準)財務危機與公司治理的高度相關,說明公司治理指標的預警效能明顯高於財務指標;在實質財務危機的判別上公司治理指標亦凸顯其相對的重要性。五、有別於其它研究使用過多變數,本研究能以邏輯化方式找出關鍵指標變數。

並列摘要


This study uses data mining algorithms to construct the company's financial crisis early-warning mode, select the Taiwan Economic Journal database (TEJ) data. To a crisis, companies with three normal ratio of 1:3, as paired samples. Samples of variable composition is divided into "financial indicators", "corporate governance indicators" and the merger of "financial and corporate governance targets" for each year for all other modes include: complete the first three years, the previous year, the first two years, status of the previous three years and do not model: real financial crisis, non-physical financial crisis. Algorithms on data mining techniques and then construct the total eighteen models of the financial crisis early-warning mode. In this study, the following conclusions: First, the back-propagation neural network (Back Propagation Network, BPN) model the best, the ability to predict the financial crisis to enhance up to 90% or more. Second, the empirical "Corporate Governance Indicator" is indeed a leading indicator, if combined with "financial indicators" and "corporate governance" indicators better. Third, the other a single year confirmed the accuracy of comparison, in addition to the previous three years' consolidated financial indicators and corporate governance indicators "slightly inferior to 0.28%, the remainder were higher than the full three years before, with the financial crisis should occur when the distance maximum (Max) due. Fourth, the model confirms the status of other non-physical (quasi-) corporate governance of financial crisis and the high correlation, indicating the effectiveness of early warning indicators of corporate governance significantly higher than the financial indicators; determine the real financial crisis on corporate governance indicators also highlight the importance of their relative of. Fifth, unlike too many variables with other studies, this study can figure out the logic of the key indicators of variables.

參考文獻


2. 黃振豐、呂紹強(2000),企業財務危機預警模式之研究-以財務及非財務因素建構,《當代會計期刊》,第1 卷,第1期:19-40。
5. 黃小玉(2004),公司治理與公司績效關條之研究一以台灣主市電子工業為例,《興國學報》,第3期:153-165。
7. 何文榮,李復禮(2006),公司治理與財務危機關係之研究-以上市公司為例,《華人前瞻研究》,第2卷,第2期:1-25。
8. 王保琪(2006),以智慧資本觀點探討台灣電子業財務危機預警模式之研究,靜宜大學,會計學系,碩士論文。
11. 林淑萍,黃劭彥,蔡昆霖(2007),企業危機預警模式之研究-DEA-DA、邏輯斯廻歸與類神經網路之應用,《會計與公司治理》,第4卷,第1期:35-56。

被引用紀錄


江明哲(2015)。資料探勘技術應用於航空業航班延誤分析-以C公司為例〔碩士論文,國立中央大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0031-0412201512045169
范睿昀(2015)。應用資料探勘技術於資源配置預測之研究-以某電腦代工支援單位為例〔碩士論文,國立中央大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0031-0412201512044755

延伸閱讀