透過您的圖書館登入
IP:3.145.58.169
  • 學位論文

運用類神經網路建構台股指數期貨預測模型

Constructing Forecast Model of Taiwan Stock Market Index Futures Based on Neural Network

指導教授 : 李維平

摘要


在經濟發展呈現全球化的趨勢和網際網路的普級帶動金融產業發展下,很多投資者不管在什麼地方都可以進行股票交易;而我們身在全球金融市場盛行的現在,一般投資者想要投資時,如果有一個精準的預測工具輔助,就能夠有信心買賣股票並且獲得高報酬。   近年來,大盤漲跌的起伏漸漸被投資人所重視,從經濟趨勢的變化可以觀察到,當一個國家的經濟開始變好時,大盤的股市就會隨之成長,而當一個國家的經濟變差時,大盤的股市則會開始衰退;由此可知,觀察大盤漲跌可以知曉經濟的發展,並且藉此使投資人了解什麼時候該進入大盤交易市場投資,而大盤通常使用技術面指標和籌碼面指標進行大盤漲跌的預測。除此之外,在國際之間金融的資訊往來越來越便利之下,從國與國之間的股市交易訊息可以看出,不同國家的金融交易互相影響股市大盤的漲跌,而為了預測更精準的大盤隔日漲跌,本研究便加入統整的國際股市指標及國際匯率指標進行隔日大盤漲跌預測的研究。   因此本研究運用能夠透過類神經網路結合多面向指標創建模型進行研究,以技術面指標、籌碼面指標、國際股市指標和國際匯率指標進行彙總,接著運用不同的訓練集時間長短進行實驗,經由實驗找出最佳參數的類神經網路模型再分別組合不同面向的指標進行研究,最後找出組合中最高預測能力的模型;主要研究是為了找出一個可以讓投資者準確判斷大盤隔日漲跌的類神經網路模型,並提供給研究者和投資人做參考。

並列摘要


The developement of the financial industry is driven by globalization and the Internet's rise in popularity. Now, investors can trade stocks wherever that are. In this globalized market, investors can trade with more confidence and profit more than usual in return with the help of an accurate tool that predicts the market.   In recent years, the bumpy stock index has been valued by investors. We can observe that the variation of stock index is depending on the country economic status. It means that we can tell the economic development status from a bumpy stock index, and the investors can understand when is the best time to invest in the stock market. It is normal to forecast the variation of stock index by technical indicators and chip indicators. In addition, the international financial information communication is more convenient, we can tell that the financial transaction between different country can influence the variation of stock index from the international stock market transaction information. To predict the variation of the stock index for the next day more accurately , the research sums up the international stock market indicators and international exchange rate indicators to predict the variation.   Therefore, a model based on the neural network and multi-oriented indicators was created from the research. By beginning with combining the technical indicators, chip indicators, international stock market indicators and international exchange rate indicators. Then conduct experiments to find out the best parameters on different training set length of time, and observed which multi-oriented indicators combination that can result in the optimal neural network model. The point is to find out the accurate neural network model to make the investors predict the stock market accurately every day, and provide the researchers and investors data for reference.

參考文獻


李天行, & 邱志洲. (2000). 類神經網路於現貨開盤指數之預測-以新加坡交易所日經 225 指數期貨為例. Asia Pacific Management Review, 5(4), 557–570.
余尚武, & 黃雅蘭. (2003). 台灣股價指數期貨套利之研究:類神經網路與灰色理論之應用. 電子商務學報, 5(2), 87–115.
洪才元. (2009). 結合基因演算類神經網路預測台股指數之模型. 中原大學資訊管理研究所學位論文, 1–95.
翁振益, & 張瑛琦. (2007). 決策分析: 方法與應用. 華泰文化事業股份有限公司.
陳淑玲, 吳安琪, & 費業勳. (2011). 臺灣股票市場技術指標之研究─不同頻率資料績效比較. 東海管理評論, 12(1S), 187–225.

延伸閱讀