具初步知識規則之分類元系統於公債殖利率預測研究

Abstract

我國一直以來，都以股票市場為發展的重點，加上股票市場的進入門檻低，各種以股票市場為標的之衍生性金融商品不斷的推陳出新，所以連帶使得股票市場參與者眾多，交易市場也十分活絡。反觀國內的公債市場，由於政府近年來努力的推動與發行公債，並且逐漸開放了各種利率的衍生性金融商品，連帶的使得國內公債市場交易成長迅速。而政府公債的買賣是以公債殖利率來報價，因此如何能夠正確的預測公債殖利率的走勢，來做出正確的投資決策是十分重要的。 而傳統上對於利率預測大多是針對總體經濟情勢的研究和分析或是以自我迴歸模型、多元迴歸模型等統計的方式來做預測，少有利用人工智慧的方法來進行債券殖利率的預測。而目前國內外，利用人工智慧的方法對於財務預測的領域上，也大都偏重在股價、指數與選擇權上，少有應用於債券相關的預測。 因此本研究嘗試應用分類元系統能夠適應動態環璄學習以及自我學習的特性，建構一個公債殖利率預測模型，進行公債殖利率的預測。本研究創新運用分類元系統在公債投資策略的研究上，為債券市場注入新的研究方法，也提供後繼有興趣的研究者一個應用分類元於公債預測上的參考以及研究的方向。

The stock market is continuously taken as the development key point in Taiwan. Moreover, stock derivatives are to weed through the old to bring forth the new, therefore the trading market is getting hot. By the government gradually opened each kind of interest rate derivatives in recent years which makes the bond market transaction growth to be rapid, therefore how to correctly forecast treasury yield rate is getting more and more important. But in the tradition, the interest rate forecasting mostly are using regression models or other statistics methods; few of them forecast interest rate by using artificial intelligence. Moreover the application of the artificial intelligence for financial forecast also mostly stresses on the stock and futures markets. It’s very few to apply on bond market. Therefore this research attempts to apply eXtended Classification System (XCS) to construct a treasury yield rate forecasting model which can adopt with the dynamic and self-learning environment, also innovatively uses XCS to help decision making on the bond investment strategy. This research provides the new research mechanism for the bond market related topics and gives successors a research reference and a research direction in the treasury yield rate forecasting.