[[alternative]]An Investment Strategy of Applying Neuro-Fuzzy into Stock Selection

碩士商業教育學系[[abstract]]本研究運用類神經模糊於台灣股票市場，研究期間自民國86年1月至民國92年12月，選定31家個股，共涵蓋11個產業，以類神經模糊進行學習並建構兩模型來預測股票報酬率。模型一以基本面變數來預測股票報酬率並與實際報酬率作比較，模型二以技術指標為輸入變數預測個股漲跌幅，並以此為依據進行模擬交易並與買入持有法及技術指標法進行比較。 研究結果發現：1、預測報酬率與實際報酬之間有顯著的正向關係，且相關係數平均皆達0.8~0.9，即以類神經模糊系統所預測的報酬率有其準確性及良好的預測效果；2、運用類神經模糊系統所建構而成的模式所獲得的投資績效報酬率均顯著的優於其他兩種投資方法，顯示出採用類神經模糊系統進行股票投資可獲得較高的投資績效；3、類神經模糊系統的獲利情形高於虧損情形，在獲利次數以及獲利平均報酬率皆有良好的表現，而利潤因素檢定值亦顯著的大於1，顯示本系統有一穩定且優良的獲利能力。[[abstract]]The purpose of this study was to construct a Neuro-fuzzy System to forecast the rate of return on investment via fundamental variables as well as technical indexes. The research period for this study was from January 1997 to December 2003. The sample of thirty-one stocks was selected from eleven industries in Taiwan. The following results have been concluded from the empirical study. First, there is a significant positive relationship between predicted rate of return by Neuro-fuzzy System and actual rate of return. In other words, Neuro-fuzzy System has a well and precise effect while predicting the rate of return. Second, the rate of return that proposed by Neuro-fuzzy System is significantly better than that of buy-and-hold strategy or technical index method. Third, the test value of profit factor analysis is grater than one and that means this system has a stable and profit-gain ability