透過您的圖書館登入
IP:18.221.145.27
  • 學位論文

運用類神經網路與時間序列分析建構台灣50股價預測模型

Forecast of Polaris Taiwan Top 50 Tracker Fund, Using Artificial Neural Network and Time Series Model

指導教授 : 邱光輝
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


股價的走勢是高度不穩定、複雜而難以預測的時間序列資料,一直以來股價預測都是學術界與投資人感興趣的話題,學者不斷嘗試透過各個策略及面向,來評估股票的價值及研判其未來走勢。 Fama (1965) 提出效率市場理論,認為市場的價格永遠處於均衡狀態,無法藉由歷史資料加以預測。但近年來,效率市場理論一直受到來自現實世界的挑戰,不理性的市場行為持續被發現。隨著行為財務學與電腦快速運算技術的崛起,技術分析與電腦模擬預測模式逐漸被學界所重視,也有越來越多學術論文證實技術分析對股價具有預測能力,可協助預測股價。而技術指標參數的決定,一直讓技術指標使用者難以抉擇,本文嘗試以技術指標回溯程式,測試不同指標參數對技術指標獲利率的影響,並找出最合適的參數組合。 時間序列的預測,傳統以自我迴歸整合移動平均模型 (ARIMA) 為主,近年來電腦計算速度的精進,使類神經網路技術又重新開始重新被重視。然而,對於複雜的股價時間序列資料,單一的預測工具常常無法得到滿意的預測結果,本文嘗試以貝氏整合結合二種以上不同預測工具,建構混合模型,以期能改善預測結果。 本研究以市場上常用之技術指標,配合台灣50指數基金之股價資料,以指標參數回測程式找出最佳參數組合,再以貝氏整合 ARIMA 與倒傳遞類神經網路的預測模型。 結果顯示,回溯機制確可有效提昇技術指標的有效性,而其所產生的交易訊號,並無法持續優於買進持有策略。但整體而言,技術分析仍有較佳的獲利能力。另一方面,混合模型,也確實能有效改善預測的結果,預測的穩定性也增強許多。

並列摘要


The movement of the stock price is highly unstable, complex, and difficult to predict. Many years, academia and investors has kept trying to assess the value of the stock and determine its future trend through various strategies and aspects. In 1965, Fama proposed the efficient market theory, said the market price is always in equilibrium and people cannot predict the stock price by using technical analysis. However, in recent years, the efficient market theory kept being challenged by the real world and the irrational market behavior continued to be found. With the development of behavioral Finance and the computing technology, technical analysis and computer prediction models have been gradually noticed by academia. Furthermore, more and more academic papers proved the predictive ability of technical analysis. Nevertheless, the decision of the technical indicator parameters is another problem. In this thesis, we use back-testing program to compute the best combination among various technical indicator parameters and to validate their profitability. Next, we use the technical indicators commonly used by the market and the hybrid model of ARIMA and BPN to simulate the trend of Polaris Taiwan top 50 tracker fund in order to improve the correctness of the prediction. The results show that overall the technical analysis can beat the buy-and-hold strategy, although occasionally buy-and-hold strategy performs the better profit. In the other hand, the back-testing program can truly increase the effectiveness of technical analysis. Moreover, the hybrid model can multiply the performance of the stock price simulation.

參考文獻


葉怡成,楊耀華,張萬鈞 (2009),「ARIMA-BPN 時間數列神經網路」,技術學刊,24卷1期:頁77-86。
Abarbanell, J. S., & Bushee, B. J. (1997). Fundamental analysis, future earnings, and stock prices. Journal of Accounting Research, 35(1), 1-25.
Abdelmouez, G., Hashem, S. R., Atiya, A. F., & El-Gamal, M. A. (2007). Neural network vs. linear models for stock market sectors forecasting. IJCNN 2007 International Joint Conference on Neural Networks, 1365-1369.
Armano, G., Marchesi, M., & Murru, A. (2005). A hybrid genetic-neural architecture for stock indexes forecasting. Information Sciences, 170(1), 3-33.
Aslanargun, A., Mammadov, M., Yazici, B., & Yolacan, S. (2007). Comparison of ARIMA, neural networks and hybrid models in time series: Tourist arrival forecasting. Journal of Statistical Computation & Simulation, 77(1), 29-53.

被引用紀錄


廖玟柔(2017)。運用類神經網路建構台股指數期貨預測模型〔碩士論文,中原大學〕。華藝線上圖書館。https://doi.org/10.6840/cycu201700811
楊子豪(2014)。供應鏈長鞭效應、敏捷力與績效關聯性之研究〔碩士論文,國立虎尾科技大學〕。華藝線上圖書館。https://doi.org/10.6827/NFU.2014.00157
陳俊佑(2013)。建構中小企業營運風險、製造策略及績效之關聯模型〔碩士論文,國立虎尾科技大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0028-1707201316221400

延伸閱讀