台股五十指數股票型基金漲跌預測之研究

Study of Taiwan 50 Stock Exchange Traded Fund Change Forecas

陳國鼎(Kuo Ting Chen)

指導教授 ： 謝俊宏

國立臺中科技大學/商管學群/事業經營研究所/碩士(2010年)

摘要

在金融海嘯過後，僅慎投資理財及小額投資已逐漸成為投資人的投資方式，在眾多金融商品中，往往太過複雜，取得資訊不易，而指數股票型基金，交易方式和股票相同，資訊透明容易取得，日漸也深受投資人青睞。傳統股票評價其預測的準確度往往取決於投資者在技術面及基本面的判斷。在本研究中，藉由有效的預測方法，探討以時間序列分析法結合倒傳遞類神網路建構「台股50隔日漲跌預測模型」，對於低點採買進策略，對於高點採賣出策略，此研究結果可提供投資大眾的參考。
本研究對象為台股50指數股票型基金。本研究資料擷取自台灣經濟新報，樣本為2007年之間台股50原始交易資料做為訓練樣本，以2008年第1季至2008年第3季的歷史資料做為預測樣本，並預測隔日台股五十指數股票型基金的漲跌。實驗結果驗証經由時間序列分析法結合倒傳遞類神經網路所建構的模型，預測隔日台股五十指數股票型基金的漲跌，機率可達七成，用以建議投資者可以單筆買入或賣出以提高其累積報酬率，因此本研究投資模型確實可提供投資人進行長期性投資時之參考依據。

關鍵字

ETF, back-propagation neural network

並列摘要

After the financial tsunami, careful and small-amount investment ways have been accepted by people gradually. For most of financial products, they are too complex and it is not easy for investors to get relative information. However, for ETF (Exchange-traded Funds), its relative information is open to get. Moreover, its trading way is same as that of stock. Due to these reasons, investors become to like this ETF gradually.

In original way, the accuracy of forecasting stock depends on investors’ judgments of technical and basic sides. In this study, combining the Time Series Analysis with Back-Propagation Neural network to make “The next day advance-decline forecast model of Taiwan ETF 50” Its strategy is buying at the low point of stock market, but selling out at the high point. The result of this study might be the reference of investors.

The object of this study is “Taiwan 50 ETF”. This data refers to “Taiwan Economic Journal”. Using original trading data of Taiwan ETF50 in 2007 as the training specimen. As for the forecasting specimen, we use the historical data from Q1 2008 to Q3 2008. Then, we use them to predict the advance and decline of Taiwan 50 ETF. The result shows the accuracy of prediction can reach to 70%, by using this“ The next day advance-decline forecast model of Taiwan ETF 50” made by the Time Series Analysis with Back-Propagation Neural network. With this way, can suggest investors to buy or sell single fund, in order to raise the Cumulative Return. Therefore, it approves that the investment model of this study can be the reference for people to make the long-term investment.

並列關鍵字

ETF, back-propagation neural network

參考文獻

蔡瑞煌 (1995)，類神經網路概論，三民書局。

林宜永 (2006)。個股股票報酬率領先落後關係之實證研究。國立中央大學財務金融研究所。

繆昆陵 (2008) 。台灣上市鋼鐵類股價指數預測之研究。國立成功大學統計研究 所碩士論文。

戴淑瑩 (2007)。臺灣50指數ETF整合型預測之研究。國立成功大學統計學系碩士論文。

蔡正修（2007）。台灣上市電子類股指數走勢預測之研究。國立成功大學統計學研究所碩士論文。