Development and Validation of a Distress Early Warning Model by Using Decision Tree Base on the Qualified Opinion
Date Issued
2014
Date
2014
Author(s)
Lee, Shu-Chuan
Abstract
This study utilizes decision tree algorithm, coordinated with financial and non-financial factors, to investigate the companies bearing qualified opinions of going-concern basis issued by CPA. Besides, this study also establishes a corporate distress prediction model by weighting variables of multi-period in order to warn of abnormal patterns of corporate financial or non-financial activities before financial distress happens.
Results from this model indicate that financial factors contain quity ratio, Interest coverage multiples, Account receivables and Days in sales of inventory and that non-financial factors contain Board pledge ratio and Related party transaction. According to the results, distress companies can be categorized into for types where more than fifty percent attributes to Type I.
Results from this model indicate that the accuracy and recall rate that correctly predicts distress companies are higher than ninety percent. In addition, based on verified results from out-of-sample and out-of-time data, this model possesses great stability without over-fitting phenomenon, showing that model tuning is unnecessary. Moreover, since this study takes Z-SCORE model developed by Altman(1968) as benchmark, excellent performance is obtained for each evaluation indicator.
Subjects
決策樹
繼續經營假設
危機預警模型
模型驗證
Z-SCORE
資料探勘
財務指標
非財務指標
Type
thesis
File(s)![Thumbnail Image]()
Loading...
Name
ntu-103-R94723091-1.pdf
Size
23.32 KB
Format
Adobe PDF
Checksum
(MD5):ce0eab45f329a37e53955ba8c1f146d8