應用市場輪廓理論的量能行為於台指期短線預測之研究

Abstract

本研究以市場輪廓(Market Profile)理論為基礎，透過成交量來取代傳統TPO建構之市場輪廓圖(TPO Chart)，並以量能與趨勢關係之依據，同時搭配技術指標，作為倒傳遞類神經網路輸入變數,以期透過物理力量的總合評判，從中萃取市場邏輯與市場結構變化之知識規則。 採用類神經網路為模型建構之人工智慧系統，配合輸入使用10單位資料轉換成市場輪廓圖指標之物理力量，進行動態時空環境下的自我學習，預測未來5及30單位趨勢方向，最後透過對預測模型的績效評估，來比較市場輪廓的量能行為與單純使用技術指標之模型間的差異性。 由實驗結果得知使用市場輪廓的量能行為指標在短線預測的趨勢方向判定優於技術分析指標，且合併兩者指標能更有效的對短線交易的趨勢方向判斷上更為顯著，證明以量能行為建構的市場輪廓指標的在短線市場有較優於傳統技術指標的預測趨勢能力，可提供投資人更明確的買賣交易資訊，以輔助決策者做正確抉擇的依據。

The research is base on theory of market profile. We can redraw the traditional TPO chart by energy index of new market profile. Depending on the Trading-Volume index and technology index, we can create the Back Propagation Neural Network, and extract the marketing rule from logic and structural variation.

By using AI system of Back Propagation Neural Network and about 10 units data of TPO chart, we can predict the tendency of future 5 to 30 units data. Finally get the difference between Trading-Volume model and technology model from analyzing the predicting result.

From our experiment, we know that Trading-Volume index is better than technology index in a very short prediction. And we can get more correct prediction by merging two of models. We have proven that using Trading-Volume index as fundamental of TPO chart can provide higher quality of marketing information instead of using traditional technology index model.