應用模糊時間序列與倒傳遞類神經網路預測台灣加權股票指數趨勢

Abstract

近年來世界各國銀行的利率均不停的走低，但是物價並沒有隨著利率一起降低，反倒是逐漸增加，在此的環境下，傳統儲蓄理財的方式已不適用，因此，該如何正確的投資理財即成為一個重要的課題。目前常見的投資理財管道有股票、債券、期貨、基金以及衍生性金融商品等，其中又以投資股票最為熱門，因此，投資散戶及專業投資經理均需要一個有效方法來預測加權股票指數之趨勢，以找出一個較佳之股票投資策略。本研究之主要目的是應用模糊時間序列(Fuzzy Time series, FTS)與倒傳遞類神經網路(Back Propagation Network, BPN)建構一個台灣加權股票指數趨勢預測模型，並利用自2007年1月2日至2010年12月31日之每日台灣加權股價指數資料來驗證本研究方法較國內外文獻所提出的預測方法有效。

For the past several years, global interest rates have gone downwards continuously, while the consumer prices have kept going upwards firmly. Under such circumstances, traditional wealth management through deposit has become useless. As a result, developing a effective novel method for wealth management has turned out to be an extremely important issue. In Taiwan, popular channels for investment and wealth management are stocks, bonds, commodities, funds, and all kinds of financial derivatives in the market. Stocks, in particular, have always been the focus of investments. Therefore, accurate anticipation of stock prices will be a conspicuous goal for which professional investment managers as well as individual investors will be craving. In this study, Fuzzy Time Series (FTS) and Back Propagation Network (BPN) are employed to construct an effective prediction model for TAIEX. Finally, TAIEX data from Jan. 2007 to Dec. 2010 are utilized to illustrate that the proposed method is more effective than the existing predictive model.