本研究嘗試透過分類迴歸樹(Classification and Regression Tree, CART),以KD隨機指標建立一交易準則的分類樹,再對每一分類以迴歸進行分析,進而偵測買進及賣出訊號,檢驗亞洲7個股票市場於07/01/1997至04/20/2005樣本期間的獲利能力。實證結果顯示,以CART所建構的交易規則於亞洲7個市場,其平均買進及賣出日報酬差異為0.348406%(相當於年報酬83.6175%),其中,台灣、香港及日本不但平均買進及賣出日報酬差異較為顯著,而且即使扣除不同水準的交易成本後,投資報酬均大於買進持有策略,顯示以CART所建構的交易規則在這三個市場具有獲利能力。
This study attempts to propose an alternative way to detect the buy and sell signals by combining classification and regression tree and KD technical indicator, and use it to examine the potential profit in seven stock markets in Asia from 07/01/1997 through 04/20/2005. Average across all seven countries and across all trading rules we evaluate, mean percentage changes in stock indices on days that the rules emit buy signal exceed means on days that rules emit sell signals by 0.348406% per day, or about 83.6175% on a annualized basis. Furthermore, Taiwan, Hong Kong, and Japan demonstrate the most consistent potential profits across trading rules, as their Buy-Sell differences are more significant and their annualized returns after various pre-specified transaction costs are larger than a simple buy and hold strategy.