整合物理量於波浪理論在股市的行為掌控

Abstract

面臨眾多衍生性商品與資訊爆發的時代，投資大眾除參考財務金融相關資訊外，更需透過各種投資輔助工具進行資訊的統整，以提高效益。如何運用適切之人工智慧方法學，建制一套可提供投資建議之輔助資訊系統，已成為不可忽視的重要議題。 技術分析應用於金融投資領域等相關研究歷史悠久。謝劍平（2004）將技術分析概分為指標類與型態類，目前已有許多技術指標應用於金融投資領域之相關研究，其中代表型態類之波浪理論相關研究數量甚少。洪若信與陳儒琦分別於2009年與2010年探討運用倒傳遞類神經網路方法學，針對以艾略特波浪特徵規則篩選後之N型波浪進行學習，研究證實經由分析型態類的物理力量，可得出股票市場中之行為規則。考量現今並無整合技術指標於波浪型態學中之相關研究，故本研究將針對技術面的技術分析，整合具重要性之技術指標及艾略特波浪理論，進行物理力量的總和評判，透過倒傳遞類神經網路的學習，使建構之模型得到優於單純使用型態類波浪理論之物理收斂結果，建立擁有更高準確率預測能力與獲利預測能力之智慧學習模型。 研究結果顯示，分別使用指數平滑異同移動平均線及相對強弱指標之技術類物理力量，整合於可代表型態類之N型波浪，運用倒傳遞類神經網路之學習，可建立優於單純使用型態類N型波浪之智慧學習模型，並提高整體準確率預測能力及獲利預測能力；另可由研究結果得知，指數平滑異同移動平均線及相對強弱指標，分別可代表股票市場中長期趨勢以及短期擾動，結合以科學化方法擷取之N型波浪後，更彰顯長、短物理能量線對股票市場具有行為掌控的能力；本研究將於得出智慧學習模型後，回溯分析技術指標之原值，進一步給予投資者投資信心之建議。

In the times of financial information explosion, if we invested the stock market with the right decision support system, we can get more profits. How to build the decision support system with the most appropriate artificial intelligence methods plays an important role. It also be one of the important issues in the financial research for these years. The research about the technical analysis has been many years. The technical analysis includes two parts, one is the technical indicators, and the other is about the trends. Now many researches about the technical indicators have been applied in the financial decision area, but there are a few researches about the wave principle of the trends. Ju-Chi Chen in 2010 indicate that extract N-wave under the Elliott Wave characteristics by using the back-propagation neural network (BPNN) method. It can have great performance in accuracy and profitability, as well as prove that stock market analysis can exactly be obtained by the physical quantity. But there were no researches about integration of technical indicators with the wave principle. For the technical indicators of the technical analysis, this thesis will through BPNN to integrate the physical quantity about the technical indicators and the wave principle. Hopefully to built one model which have the better accuracy predictions and profitability predictions than the model only with the wave principle. The results showed that through scientific method to integrate RSI and MACD with the N-wave will have the better accuracy and profitability predictions than only considering N-wave physical quantity. Considering that RSI represents the short-term disturbance and MACD represents the long-term trends, the results manifested that long-term and short-term physical quantities also have the control ability on the stock market. After establishing an intelligent model, this thesis will retrospect indicators original value, gives more confidence recommendations to investors.