應用分類元系統於台指選擇權日內價格分析之研究

Abstract

本研究應用分類元系統進行指數選擇權日內交易價格分析研究。諸多實證研究指出台灣股市為淺碟式經濟體。而進行如日內極短線交易之短期投資操作，相較於長期持有投資操作，其正具有可適當規避該環境下的高度系統風險特點。同時，指數選擇權商品所具有的避險能力及相對偏高的價格漲跌幅度，適足以提供投資人獲取低風險高報酬的機會。是以本研究即以指數選擇權商品為本研究的標的物，嘗試應用具有動態學習能力之分類元系統，透過對指數選擇權日內交易之歷史資料進行學習，建構一具有可適當規避系統風險能力之短期避險預測系統。 本研究以Wilson XCS系統架構為核心，並從B-S評價模式中選取輸出入因子。在該些因子，本研究模式同時考量期貨價格及成交量與隱含波動率變化。進而選取台指選擇權為本研究模式進行短期避險預測之金融商品，模式分別有買權模式及賣權模式。至於在模式效能評估上，本研究以隨機模式為準確率評估之對照組。而在累積獲利評估部分，本研究則以Buy and Hold操作策略為對照組。實證指出，本研究模式經由XCS動態學習過程歸納選擇權市場的日內交易行為，同時針對次十分鐘與次二十鐘的權利金變化幅度之預測的準確率與其投資累積獲利模擬均較對照組有顯著之結果。

This study applied learning classifier system on Index Option intraday trading analysis. Many researches have showed that Taiwan stock market is a highly systematic risky economy. Investors have to adequately avoid the highly systematic risk to accumulate the potential profit in this kind of economic environment. By the way, the short term investment, such as intraday trading, is compared to the long term investment, is a well-chosen strategy to investors accumulating the profit. Owing to the Index Option is a kind of hedging financial commodity and the related higher range of price fluctuation, investors could bring it up and have the opportunity to obtain the higher profit with lower risk. Therefore, this study tried to build up an Index Option intraday trading model that gives the investments by holding short term position to avoid systematic risk. As for the model kernel, this study applied learning classifier system, an integrated artificial intelligence approach that incorporates reinforcement machine learning method to Genetic Algorithm with the dynamic learning ability. And the output of this model would be a suggestion of the trend prediction of short term price fluctuation to investors.

Furthermore, this study utilizes Wilson’s XCS as the system kernel of this model. The input and output formation of this model is drawn up according to B-S pricing model. By the way, the input factors of this model also are concerned with the moving of the Future price, volume, the Index Option volume and its implied volatility. In the Experiments, this study selects Taiwan Index Option as the commodity to the prediction system with two parts, call-option and put-option. To evaluate the option model performance, this study is respectively compared the accuracy with the Random Walk Model and compared the accumulative profit with Buy-and-Hold strategy as well. Finally, the empirical result shows that the system could be successfully generalized the Index Option intraday trading rules of short term prediction with 10-minute and 20-minute, and all of them have been exhibited remarkable outcome.