臺灣新上市櫃公司失敗風險預測;Predicting IPO failures in Taiwan stock market

NCUIR > School of Management at National Central University > Graduate Institute of Finance > Electronic Thesis & Dissertation > Item 987654321/83738

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/83738

Title:	臺灣新上市櫃公司失敗風險預測;Predicting IPO failures in Taiwan stock market
Authors:	許博惇 Hsu, Bo-Dun
Keywords:	新上市上櫃公司;機器學習;興櫃市場;IPO 失敗風險;IPO;Machine learning;Taiwan’s emerging stock market;IPO failure risk
Date:	2020-07-08
Issue Date:	2020-09-02 16:59:54 (UTC+8)
Publisher:	國立中央大學
Abstract:	本文透過IPO特性變數以及興櫃市場相關變數預測IPO公司是否會於上市(櫃)後五年內下市(櫃)或轉為全額交割股。研究後發現興櫃市場相關變數的確為預測IPO 是否失敗之重要變數，加入興櫃市場變數的確能提高樣本外預測準確度，且亦能增加羅吉斯迴歸(logistic regression) 之Pseudo ??2。此外，本文使用羅吉斯迴歸、隨機森林 (random forest) 與extreme gradient boosting 方法建構模型，發現以台灣2002-2014 年預測IPO五年內失敗可能性來說，羅吉斯迴歸樣本外預測效果最佳，而模型穩定度最好的則是extreme gradient boosting 方法。;This thesis constructs an IPO failure prediction model with IPO characteristics and the trading or financial information in Taiwan’s emerging stock market. The results indicate that the trading or financial information in Taiwan’s emerging stock market are important and significant variables to predict IPO failure risk. We find that AUC and Pseudo ??2 increase after including the variables related to Taiwan’s emerging stock market in the prediction model. In addition, we build the prediction model by logistic regression, random forest and extreme gradient boosting to predict whether IPO will fail or not in Taiwan from 2002 to 2014. The results indicate that logistic regression performs better than random forest or extreme gradient boosting in predicting IPO failures. Finally, we find that the performance of extreme gradient boosting is the most stable in the repeated cross validation.
Appears in Collections:	[Graduate Institute of Finance] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	196	View/Open

View Licence

社群 sharing

Loading...