運用市場輪廓動量效應於台指期市場之行為發現

Abstract

金融市場是一個充滿變化的動態環境，而市場的價格變動也受到各種因素所影響，由於市場價格變幻莫測，因此許多人開始認為市場的價格走勢服從隨機漫步理論。Steidlmayer提出市場輪廓理論，反駁隨機漫步之觀點，主張市場價格之變化是由於市場參與者之交易行為而變動，認為市場具有其行為規則。 本研究基於市場輪廓理論之擺動因子，提出一趨勢指標衡量市場長線之價值動量變化，試圖運用價值變化的動量效應以提升隔日之短線交易績效。在當日短線交易中，使用隨機指標(KD)、指數平滑異同移動平均線(MACD)與資金流量指標(MFI)及定量市場輪廓指標評估台指期當日價格走勢。本研究將上述指標作為倒傳遞類神經網路輸入變數，透過類神經網路之非線性學習能力來辨識臺灣指數期貨市場之行為知識。 由實驗結果可知，衡量市場長線動量趨勢變化，有效提高了模型之預測準確率及獲利能力，顯示市場價值動能趨勢對於當日短線之價格變化有其重要性。另外，將本研究之實驗組與隨機交易做比較，發現使用市場輪廓及技術分析之模型較隨機交易更能從市場中獲得利潤，亦表示台灣加權指數期貨市場不服從隨機漫步理論。

Financial market is a dynamic environment full of changes, and the market price changes also affected by various factors. Due to the vagaries of market prices, many people believed that the market price movements follow random walk. Steidlmayer proposed market profile theory to refute the view of the random walk. He believes that changes in market prices is due to the trading behavior of market participants so the market has its logical rules. This research establishes an indicator based on rotation factor in market profile theory to measure the momentum of market value in the past few days. The propose of this research is integrating the information about the trend of past few day's value area into five minute information in a day to improve the model which only use five minute information. In the day trading, we use stochastic (KD), moving average convergence-divergence (MACD), money flow index(MFI) and quantitative market profile data to evaluate the direction of price changes in one day. We use above indicators for the input of back propagation network to discover the knowledge of Taiwan index future markets by its nonlinear data processing ability. The results of experiments indicates that using the value area rotation factor to measure the momentum of value area in past few days improves the accuracy and profit of the predict model. It implies that the trend of momentum in the past few days is important for predicting the direction of price in one day. In addition, we compare the experimental group with random trading, and we find our model could get more profit than random trading. Therefore, it indicates Taiwan future market doesn't follow random walk.