標題: 以日構成市場輪廓於股價行為研究---以台灣上市公司為例
Research on stock price behavior by constituting Market Profiles in days-Evidence from Listed Companies in Taiwan
作者: 彭嘉彥
Peng, Chia-Yen
陳安斌
Chen, An-Pin
管理學院資訊管理學程
關鍵字: 市場輪廓;倒傳遞類神經網路;台灣50成分股;振盪因子;Market Profile;Neural Network;TSEC Taiwan 50 index;Rotation Factor
公開日期: 2012
摘要: 根據台灣證劵交易所統計資料顯示,至2012年為止,國內股市投資累計開戶數為16,546,146戶,且總成交值239,003.6億元,證券交易量值占了舉足輕重的地位,顯示股票為國內投資人最主要理財工具。近年來實證研究指出,股市行為存在特定型態而具有可預測性,有許多現象的發生,為傳統財務學所無法解釋。技術分析專家相信在歷史價量中存在著一定的行為模式,可用來預測未來的價格走勢。 本研究以市場輪廓理論為基礎結合價格波動的趨勢擺動因素,藉由類神經網路學習,從價格與價值間關係(價格偏離值)和價格波動擺動趨勢(振盪因子)中萃取市場邏輯與市場結構變化之知識規則,觀察市場輪廓所產生之資訊,對台股現貨市場未來漲跌趨勢評估及區間的預測能力,發掘金融投資環境中的知識及行為模式。 由實證結果得知,本研究所提出的整合市場輪廓理論的類神經網路預測模型,在股價趨勢的預測上,與隨機交易模型相比較,整合市場輪廓理論的類神經網路預測模型不論是在預測股價漲跌準確度上或是平均獲利能力上均明顯較佳。證實了股價趨勢是可以掌握與預測的,同時也說明了股市價格走勢並不符合隨機漫步假說,市場不具效率性。 投資者可利用市場輪廓分析搭配適當的交易策略來增加投資獲利的機會;證明市場輪廓理論所蘊含的知識規則是有其效用的。
In recent years, much research indicates that there is some specific market behavior and patterns can not be explained. Technical analysts believe that the stock price and volume in the history that can be used to predict future price. This study based on the Market Profile theory implementing neural network architecture with market profile price indicator and rotation factors. Attempting to explore the knowledge and behavior patterns on the Taiwan stock market trend assessment and forecasting interval in financial investment environment Comparing with the random trading neural network model , the experimental results show whether accuracy in predicting the ups and downs on the stock price or profitability has significantly better results. Demonstrated price trends can be predicted, and shows the stock price trend does not meet the Random Walk. Investors can take advantage of market profile analysis with the appropriate trading strategy to increase the opportunities for profitable investment.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT070063401
http://hdl.handle.net/11536/71821
Appears in Collections:Thesis