應用資料探勘探討共同基金績效之持續性與從眾行為

The Study of Using Data Mining on the Mutual Fund Performance and Herd Behaviors

"共同基金是由具備專業知識與背景的經理人操盤，當同時面臨著績效壓力與競爭對手等因素，可能會同時買入或賣出某些股票，即稱為從眾行為。本研究採用資料採礦的研究方法，以國內股票開放型一般股票型基金共106支為研究對象，首先探討國內股票型基金的績效表現是否具有持續性，接著再利用各基金的持股明細探討基金經理人的從眾行為。 研究結果顯示，短期績效表現佳之基金在長期亦會有好的績效表現，而短期績效表現較差的基金，長期的績效表現也會較差，說明了基金的績效具有持續性。而從基金持股明細的資料探勘結果中發現，在績效表現較佳的基金中，基金經理人持有聯發科(2454)同時亦會持有台積電(2330)。但在績效表現較差的基金中，卻無法發現基金經理人有同時持有某些股票之從眾行為。"

"Mutual funds are traded by managers with professional knowledge and background. When faced performance pressure and competition, they may buy or sell certain stocks at the same time, which is called herd behavior. This research adopts data mining as the research method, taking 106 domestic open end equity funds as the research object, first discusses the performance persistence of funds, and then uses the stock holding details of each fund to explore herd behaviors of fund managers. The results found that funds with good short-term performance will also have good performance in the long-term, while funds with poor short-term performance will have poor performance in the long-term, indicating that domestic open end equity funds exist performance persistence. The results of data mining and fund's shareholding details show that a fund manager with better performance holds MediaTek (2454) and TSMC (2330). However, in funds with poor performance, it is impossible to find that fund managers have the relevance of holding certain stocks at the same time."