標題: 應用多重類神經網路於台灣期貨指數極短線走勢行為知識發現
Applying Multi-Neural Network on Short Term Intraday Trading of Taiwan Index Futures Market
作者: 許惠喬
陳安斌
資訊管理研究所
關鍵字: 多重類神經;日內交易;台灣指數期貨;技術分析;Multi-Neural Network;Intraday Trading;TAIEX Futures;Technical Analysis
公開日期: 2010
摘要: 台灣是一個淺蝶式市場,股市易受消息面影響而大幅波動,同時美股與台股的連動性高,隔夜的風險使得投資人長期獲利在一夕之間縮減,採用日內投資交易可規避隔夜持有之風險。但自從當沖保證金減半制度實施、政府連續調降交易稅、電子交易使得手續費逐年下降,使得期貨交易成本大幅降低,日內投資參與者與日遽增,當沖獲利空間增加的同時亦帶來風險;因此本研究為了規避這些風險,加入了長線保護短線的概念,以輔助投資人進行日內交易的決策擬定。 本研究提出以多重類神經網路為架構,搭配長短期技術分析,學習台灣加權指數期貨日內趨勢行為,嘗試從股價趨勢行為中,找出知識規則。運用多重類神經網路來針對長期、短期單一網路做總和評判,使類神經網路的輸出更具有可靠性,建立一個預測日內極短線指數趨勢的預測模型 由實驗結果得知,多重類神經網路模型在預測能力以及獲利能力上,表現較單一類神經網路模型優異,準確率提升。由此可知多重網路經總和批判,統整長、短期物理力量後的效果確實會優於單一網路。同時也證實了藉由長線保護短線的概念來進行日內極短線的投資操作,可以有效降低日內股價波動的風險。
The stock market in Taiwan is a shallow-plate market, which is often vulnerable to sharp fluctuations by news side effects. Besides, U.S. stock markets and Taiwan stock index are highly correlated. Investors may lose their long-term profits quickly due to overnight risks, therefore intraday trading can be used to avoid such risks. However, since the intraday trading futures margins reduced by half, futures transaction tax reductions, and decreased electronic transactions fees year by year, these factors increase the intraday trading investors and also reduce the futures transaction cost substantially. Increasing in daily trading profits also increase the risks. Therefore, in order to avoid these risks mentioned above, this study adds the concept of long-term protection of short-term to assist investors on intraday trading decisions. This study proposes a multi-neural network model with long-term and short-term technical analysis and tries to find the knowledge rules of the trends in TAIEX Futures’ intraday trading behavior. By using multi-neural network, we make integrated evaluation of long-term and short-term subnetwork, and verify the more reliability of the neural network’s output. Therefore, a very short-term intraday trading of Taiwan Index Futures trend forecast model is established. The results show that multi-neural network is significantly more effective than single neural network model in forecasting accuracy and trading profitability. We also confirm that the concept of long-term protection of short-term can effectively reduce the risk of intraday trading stock price volatility.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079734516
http://hdl.handle.net/11536/45481
顯示於類別:畢業論文


文件中的檔案:

  1. 451601.pdf
  2. 451601.pdf