在技術分析的眾多技術指標中,由於K線圖可以簡單的圖形化方式反應市場的趨勢且投資人可以藉由K線型態的成立與否來判斷市場未來走向,所以常被投資大眾所採用。然而,不同的投資人對於同一個K線型態的解讀往往不同,且K線型態資料的層級性更加深了辨別K線型態的困難度。因此本研究以自組織映射圖網路模式及成長型階層式自組織映射圖網路模式分別建立一套以K線型態辨識為基礎的投資決策系統,研究對象包含台灣指數期貨、台灣金融指數期貨、台灣電子指數期貨、美國道瓊指數及美國那斯達克指數等五種金融商品的歷史資料。實證結果顯示,成長型階層式自組織映射圖網路模式除了可探勘出隱藏在資料中的重要型態,且在多項指標上的表現皆優於自組織映射圖網路模式,可協助投資者獲得更高的超額報酬。
Among all technical analysis methods in stock market, K chart is widely adopted by investors because of its friendly usage and graphic display of market trend. However, K chart patterns are usually differently identified by investors, and the hierarchical characteristic of K chart patterns also seriously confuses investors. As a result, an investment decision support system based on growing hierarchical self-organizing map is proposed in this paper to help investors properly distinguish different K chart patterns. The experimental results reveal that growing hierarchical self-organizing map is better than self-organizing map for K chart pattern-recognition in correlated process data.