透過您的圖書館登入
IP:3.135.216.174
  • 學位論文

運用資料探勘技術建構共同基金合併清算預測模型

On the development of a Forecasting Model for Liquidation and Merger of Mutual Funds

指導教授 : 劉士豪

摘要


目前於各個投信公司網站或是基金資訊源網站都可查得各共同基金的相關資料,包括短期、中期、長期的基金績效或是投資、波動風險,但是目前並未有共同基金清算合併的預測服務,且尚未有學者針對共同基金之清算合併預測進行研究。 本研究透過資料探勘技術,進行共同基金的清算合併預測研究,並依投資學上短期投資定義,針對一年內的時間點分別建立預測模型,以歸納共同基金於清算或被合併當時、於清算或被合併前三個月、於清算或被合併前六個月、或於清算被合併前一年之特徵,以利投資人做出適當的投資決策;本研究所採用的資料探勘技術分別為叢聚分析、決策樹分類;叢聚分析主要是判斷本研究的各個時間點下,已遭清算或是已被合併的共同基金是否有明確的特徵可以區分,亦或皆有相同的警訊,決策樹則是用來建立本研究的預測模型。 本研究透過叢聚分析實證只有在清算合併當時的分群特徵較明顯,在其他幾個研究時點,被清算合併之基金無法明確分群;而本研究之預測模型亦根據叢聚分析之結果,決定每個研究時間點的輸出參數,各個研究時間點的預測結果以共同基金發生清算合併當時的預測效果最佳,但除清算合併當時的研究時間點外,其他時間點的預測Response Rate都可達100%,表示本模型在其它三個研究時間點,預測該共同基金會發生被清算或被合併皆準確,但此三個研究時間點下,Recall Rate約在7成上下,表示預測模型在共同基金清算合併的預測能力約7成。

關鍵字

決策樹 叢聚分析 共同基金 清算 合併

並列摘要


All of the Fund houses periodically publish the information of mutual funds on the web sites. The information includes both performance and risk in short-term、middle-term and long-term . But there isn’t any mutual funds liquidation or merger prediction be disclosed, there are few studies about forecasting for liquidation or merger of mutual funds, either. The study intends to develop the forecasting model for liquidation or merger of mutual funds by data mining. Base on the definition of short-term in financial, the study develops four kinds of forecasting model with different period in one year. The first one is the day of mutual fund liquidation or merger which named “the day model”, the second one is three months before the day of liquidation or merger which named “the 3-months model”, the third one is six months before the day of liquidation or merger which named “the 6-months model”, the last one is twelve months before the day of liquidation or merger which named “the 12-months model. In each models, the study determines if there are any obvious different features between the liquidated mutual funds and the merged mutual funds by Clustering. The output parameters of the four forecasting models are based on the results of the four clustering models. The forecasting models construct from decision tree to predict the liquidation or merger mutual funds in short-term rules of liquidation or merger of mutual funds. Except the day model, the liquidation funds and merger funds are clustered together in all the clustering models. The Recall rates of all the decision tree forecasting models are around 70%. The response values of all the decision tree forecasting models are 100%, except the day model. As the results, the study can be used to forecast the liquidation and merger of mutual funds.

並列關鍵字

Liquidation Mutual Fund Decision Tree Clustering Merger

參考文獻


8. 陳生祥. 2005. 運用資料探勘技術建構企業財務危機預警模式-結合財務與非財務資料. Ph.D. diss., 中原大學; 資訊管理研究所.
11. 林志娟, 江妙真, 張慶暉, 溫博仕. 2005/06/01. 平衡型共同基金績效評估與擇時能力研究. 智慧科技與應用統計學報 3, no. 1:109-131.
2. 王堅斌. 2003. 使用分類結合法與決策樹判別忠誠顧客. Ph.D. diss., 國立成功大學; 資訊管理研究所.
3. 章碩麟. 2004. 美台共同基金交易弊端之探討. Ph.D. diss., 臺灣大學; 財務金融學研究所.
5. 謝邦昌. 2005. 資料採礦與商業智慧-SQL Server 2005 Data Mining & Business Intelligence with SQL Server 2005.初版 ed. 臺北市: 鼎茂圖書出版股份有限公司.

延伸閱讀