在股票分析的實證研究上，大多以基本面分析、技術分析為工具，或藉由探討外資、投信及自營商三大法人間的資訊領先地位，試圖從中找尋出最佳的投資時點，但是這些研究卻未對個股籌碼變化做深入探討，因此個股籌碼流動向量對股價的影響如何，乃本研究之探討重點。本研究運用倒傳遞類神經網路，研究在交易的過程中，個股籌碼流動向量與股價波動關係。研究資料為台灣證券交易所個股交易日資料，其中2003/10/01~2004/08/31的資料為訓練測試期所使用，2005/07/20~2005/10/28的資料為驗證期所使用。本研究以台積電建構個股籌碼流動向量股價預測系統後，以該系統驗證台指50成分股的適用範圍，實證結果如下： 一、以倒傳遞類神經網路不同的參數設定值經過多次的測試後，定位出各項參數較佳的設定值，所建構的預測系統確實可對股價漲跌進行預測。二、本研究所建立的個股籌碼流動向量股價預測系統可對於樣本在不同期間的股價進行預測。三、本研究所建立的個股籌碼流動向量股價預測系統可對台指50成分股不同個股的股價進行預測。 Most previous empirical studies seek to discover the opportune moment of stock investment mainly through the fundamental analysis, technical analysis and the leading information from the three key institutional investors (foreign institutional investors, domestic mutual funds and security dealers). However, these studies neglect the vibrations in the stock pricing by the trading vector, which could also shed light on the change of stock price. The purpose of this study is to fill this void and focus on the relationship between trading vector and price change in the Taiwan’s stock market. The Back-propagation Network (BPN) is used to establish the price prediction system. Firstly, we use daily data of Taiwan’s stock on the Taiwan stock Exchange from Oct. 01, 2003 to Aug. 31, 2004 to train and test the prediction system. Secondly, the daily data during the period July 20, 2005 to Oct. 28, 2005 are used to verify the prediction system. After establishing an initial system for the Taiwan Semiconductor Manufacturing Company Limited (TSMC), the prediction system is employed to verify all of the stocks in the Exchange Traded Funds (ETF). The conclusions are as follows:1.After test for different parameters in BPN for many times, we got some suitable digits to establish the prediction system, and this system can be used for prediction.2.The price prediction system can be used for different periods.3.The price prediction system can be used for different individual equity securities in the ETF.
