整合擺動因子與波浪理論於臺灣期貨指數走勢行為研究

Abstract

面對瞬息萬變的股市，投資人常因無法掌握市場方向而蒙受重大損失，因此如何有效分析與判斷股價變動的趨勢便成為一項重要的議題。本研究基於艾略特波浪理論，並結合市場輪廓理論中的擺動因子，透過倒傳遞類神經網路進行總和評判，以更為凸顯股市輪動時所隱含的物理力量，進而掌握股價波動的行為模式。 針對擺動因子的部分，除了基本定義的計算方式(RF)外，本研究亦額外提出三種計算方式，分別為考量各時間區間內買賣雙方的整體競價活動(RF_M1)、考量各時間區間內買賣方實際的攻擊力道(RF_M2)以及考量買賣方實際的攻擊力道與整體競價活動(RF_M3)，期能更為有效凸顯該值所隱含的物理力量。 實驗結果顯示本研究所採用的擺動因子計算方式中，結合N型波浪後，以RF擁有最佳的預測效果。若進一步同時採用兩種計算方式，則以RF搭配RF_M3之組合展現最佳的績效成果，優於僅使用基本計算方式之實驗組。而以N型波浪的作為基礎，並同時納入擺動因子進行整合後，相較於單獨使用N型波浪作為交易策略的依據，皆有著更高的準確率與獲利能力，表示擺動因子的加入確可凸顯各波段所隱含的物理力量，並強化其趨勢性。

Investors often can't catch direction when facing the rapidly changing stock market. Therefore, how to analyze and determine the market's trend becomes an important issue. Based on the Elliott Wave Theory, this study further tries to include the rotation factor mentioned in the theory of market profile. Then, the back-propagation neural network is used to integrate the physical quantity about the wave principle and rotation factor in order to manifest the physical quantity and catch market behavior behind market's rotation. For the calculation of rotation factor, besides the basic method(RF), this study extra proposes three methods for considering the total auction process between buyers and sellers in each time period(RF_M1), and for considering the actual buying and selling power in each time period(RF_M2), and for considering both the actual buying and selling power and the total auction process in each time period(RF_M3). The result shows that after combined with the wave principle, using RF to calculate rotation factor is better in prediction than other methods. Further, using the combination of RF and RF_M3 has better performance than using RF only. And after the integration of the wave principle and rotation factor, it is better both on accuracy and profitability than the model only with the wave principle, showing that the use of rotation factor in the wave principle can further manifest the physical quantity and enhance its attempted direction.