應用延伸性分類元系統預測利率期貨價格走勢-以十年期公債期貨為例

Abstract

據FIA調查顯示，全球全十大衍生性金融商品中有七個為利率的衍生性商品，利率衍生性商品之重要性由此可見一般。由於近年來許多學者應用人工智慧於解決財務金融問題上，均得到不錯之成果。故本研究應用延伸性分類元系統（eXtended Classifier System, XCS），採用技術指標作為輸入因子，以隔日之價格走勢做為輸出，建置一個可對利率期貨進行價格走勢分析之輔助決策系統，並設計一套交易機制，依照模型建議之動作進行樣本外模擬交易測試。

實證結果顯示此模型之預測準確率與報酬率均優於隨機買入與傳統投資策略，在不同的時空環境下，亦具有適應性與平穩之預測性。故本研究結論為，採用具動態環境學習之延伸性分類元系統，結合技術指標，可有效掌握利率期貨之價格走勢，適於作為投資者之決策輔助系統。

According to statistics from the Futures Industry Association (FIA), there are seven interest futures and options out of globally top 10 derivatives, so interest derivatives are essential financial products. In recent years, many scholars used A.I. (artificial intelligence) on the financial problems, and problems were solved effectively. In this study, eXtended Classifier System (XCS) is applied to develop a model, in which technical indicators are taken as input factors to predict next day's price trend. To prove that XCS could be applied to forecast interest futures, the performance of XCS is compared with that of random work and of trend-following trading strategies.

The simulation and statistical results show that XCS-based model has more precise forecasting accuracy and better profit-earning capability than other comparison models. This system can also suit in various environments, and provide stable predictions. Thus, this system is qualified to be the decision support system for investors.