以GA-SVM法探討企業財務危機之研究

Exploring Financial Distress Based on GA-SVM Method

邱登裕(Deng-Yiv Chiu) ； 鍾典村(Tien-Tsun Chung) ； 吳致遠(Chih-Yuan Wu) ； 謝齊莊(Chi-Choang Hsieh)

《中華管理學報》 8卷4期 (2007/12) Pp. 61-85

https://doi.org/10.30053/CHJM.200712.0004

摘要

長久以來，企業財務危機預警模式的研究一直是政府機關、金融業者、企業單位及投資者所關注的課題，然而傳統上的相關研究大多偏重以財務變數來建構模式，而忽略非財務變數。財務報表的數字往往代表的是企業經營結果的呈現，當財報數字公佈時，投資人才知悉公司狀況而想出脫手中持股，往往為時晚，熟知公司內情的內部人可能早在事件爆發前出脫持股。若能預先知悉內部人持股及其它非財務變數的變化，也許對財務危機預警有所幫助，因此本研究同時以公司治理及財務指標探討企業財務危機預警模式之關鍵因素為何。
本研究以財務變數與非財務變數作為輸入變數，先使用基因演算法(GA)挑選出具影響的變數，再使用支持向量機(SVM)進行探討，並利用建構好的GA-SVM架構進行階段式分析及多年預測分析，以分析相關變數對整體企業財務危機的影響性。本研究提供了一個不同的財務危機預警模式，實驗結果發現，二階段危機預警模式優於單一階段預警模式，且多年預測分析方法有較高的凖確率。

關鍵字

企業財務危機 ； 公司治理 ； 基因演算法 ； 支持向量機

並列摘要

For a long time, financial distress is a critical research for government organizations, financiers, enterprises and investors However, the relevant researches traditionally emphasize the financial variables for building the model mainly. The financial statements often show the results of enterprise management. It is often too late for investors to transact their shares after the financial statements are announced. But for investors familiar with inside corporate governance, they can take proper actions before financial distress breaks out. In this research, we explore the critical factors of business financial distress in view of corporate governance. The factors in study include corporate governance and financial statement. Therefore, this research tries to explore the key factors of financial distress in view of corporate governance and financial indicators.
First, in stage-analysis method, we use financial variables and non-financial variables as input variables. Secondly, we select the influence variables by using genetic algorithm. Thirdly, we perform classification by using support vector machine. Finally, we construct two-stage business financial distress model and multi-year analysis model. And we explore the influence of variables on the corporations.
This research provides an alternative approach for financial distress analysis. We find that both of corporate governance and financial indicator are key factors to be considered. The empirical results show that two-stage model outperforms one-stage model in distress classification. And, multi-year analysis model has better performance. The proposed model could provide valuable reference for financial distress.

並列關鍵字

Financial distress ； Corporate governance ； Genetic algorithm ； Support vector machine