標題: 一個企業財務危機預測的整合性方法
An Integrated Approach for Corporate Financial Failure Prediction
作者: 游濬遠
Yu, Chun-Yuan
丁承
Ding, Cherng G.
經營管理研究所
關鍵字: 財務危機預測方法;成長型層級性自映射網路圖模式;分類迴歸樹模式;軌跡分析;Financial Failure Prediction;Growing Hierarchical Self-organizing Map;Classification and Regression Tree;Trajectory Analysis
公開日期: 2015
摘要: 對於投資人或是債權人而言,企業無預警地違約除了對銀行帶來呆帳之外,亦會對投資人或債權人帶來巨大的損失。因此如何建立一個穩健的財務危機預測模式以輔助投資人及債權人進行決策,便成為極重要的議題。事實上,過去學者已經提出許多的財務危機預測模式,然而,這些模式在發生危機前兩年或三年的預測績效都會大幅降低。因此,本研究目的為建構一個財務危機預測的整合性方法,使投資人在進行短期或長期財務危機預測時得以獲得優於傳統方法的預測正確率。此外,並提供一簡單的圖形化方式讓投資人了解企業整體財務概況演變。 本方法結合成長型層級性自映射網路模式、分類迴歸樹方法以及軌跡分析等三種模式。其中,前兩種方法被用以建構財務危機預測模式,而軌跡分析則被用以了解企業財務狀況的改變。為了驗證及衡量本研究所提的整合性方法,本研究以台灣上市櫃公司作為實證樣本進行分析。實證結果發現,本研究所提出之方法在財務危機發生前一年、前兩年、以及前三年的預測上皆有良好的預測能力,其預測能力優於傳統常用的標竿模式 (鑑別分析模式、羅吉斯模式、倒傳遞類神經網路、多元適應性雲形迴歸及支援向量機等)。此外,分類迴歸樹所產生的規則更可讓投資人掌握企業發生財務危機的關鍵資訊。因此,建議投資人可以使用此方法來進行企業財務危機預測。
Predicting financial failure of a company is important for an investor or a creditor because financial failure without any symptom not only results in banks’ bad debts, but also causes huge losses to investors. Although there exist many models for predicting financial failure, their performance over a period of two or more years is poor. Therefore, the purpose of the present study is to propose an integrated approach by which a relatively higher accuracy rate, regardless of short term or long term prediction, can be obtained and the dynamic change of a company's financial status can be captured by visualization. The approach combines the growing hierarchical self-organizing map (GHSOM), the classification and regression tree (CART) and the trajectory analysis. GHSOM and CART are used for prediction and trajectory analysis for visualizing the financial change. To demonstrate and evaluate the proposed approach, we collected data of healthy and failed companies sampled from the Taiwan stock market. The empirical results indicate that the proposed approach outperforms five commonly used benchmark models (discriminant analysis, logistic regression, back-propagation network, multivariate adaptive regression splines and support vector machine) in one-year, two-year and three-year ahead predictions and provides insightful information and therefore it is recommended for financial failure prediction.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079437811
http://hdl.handle.net/11536/126829
Appears in Collections:Thesis