破產預測模型之比較

Abstract

公司的倒閉風險，在學術界及銀行界都是相當受到重視的議題，一個判斷上的失誤，可能會產生重大的影響。本論文希望可以利用各種不同的資料分類方法，從傳統的Z-score、Logit迴歸到近期相當熱門的類神經網路及支援向量機等方法，尋找出較佳的破產預測模型，使企業或投資人能夠在公司營運出現問題前及早做出因應，以避免產生巨大的損失。

The risk of corporate bankruptcy has been an important topic for the academia and the banking sector, as any misjudgment may lead to disasters. This thesis compares the performance of various data classification methods, such as Z-score, Logit regression, neutral networks and support vector machines. With a better model, corporations and investors can be better prepared before companies show signs of bankruptcy so that huge losses may be avoided.