運用類神經網咯於臺灣股票市場價量關係的預測與分析

Abstract

股票價格和成交量是了解股票市場結構必要的基礎。為此，本研究即是針對臺灣股票市場中，股價指數和成交量之間的變化關係進行預測與分析，其中是以目前的價量關係，配合技術分析的方法，並透過類神經網路的技術，以掌握到價和量的行為模式，同時也可以藉此進一步了解股市未來發展的趨勢。 在本研究中，首先收集民國78年1月到民國83年1月的股市日資料和週資料，再依照股價和成交量的漲跌幅度，將資料分成價漲量增、價漲量縮、價跌量增四種群組。然後針對每個資料群組，分別以四種不同的資料輸入形態，配合兩種類神經網路模式，以進行網路的學習與測試。實驗結果發現，將資料依照價量關係分組後的預測結果，其準確度多會優於資料未經分組的預測結果，說明了具有相同走勢的資料，容易具有相同的漲跌特性。而且無論是在日指數或是週指數方面，對於未來價量趨勢預測都可達到70%以上的準確率。顯示本研究所建立的價量預測模式已具有初步的預測能力。

Stock prices and transaction volumes are the necessary bases for understanding the structure of the stock market. Taking this as given, the aim of this study is to forecast and analyze the relation between the stock price and the volume in the Taiwan Stock Market. In this study, the current price-volume relationship is combined with the technical analysis method and artificial neural network to capture the behavior pattern of the stock price and the volume. With this, the trend of the stock market can be understood more clearly. This study has used the daily data and weekly data of the Taiwan Stock Market from Jan. 1989 to Jan. 1994. According to the rising-falling range of the stock price and the volume, this data is divided into four groups. For each data group, four types of input data are learned by two neural network models. The results show that using the divided data to forecast will be more accurate than using the undivided data. That is to say, if the data has the same trend, it will have the same fluctuating characteristics. And for daily data or weekly data, the accurcy of forecasting the price or volume trend can reach over 70%. This might suggest that the model proposed by this study has a forecasting ability.