透過您的圖書館登入
IP:3.17.79.60
  • 學位論文

股市關鍵技術指標萃取於智慧型交易系統之研究

Extraction of key technical indicators in intelligent stock trading system

指導教授 : 張百棧

摘要


對於所有預測問題來說,特徵的使用,都是不可或缺的,而如何去挑選合適的特徵,成為一個問題,究竟在眾多的特徵之中,我們必須如何去做出篩選,以符合問題所需,達到維度縮減的目的。 本文提出了一種基於核函數主成份分析(Kernel Principal Component Analysis)中提取關鍵特性,以提高股票交易模型的性能。核函數主成份分析是一個基於核函數的數據映射,所以它同時具有PCA和非線性映射的特性。許多研究均採用逐步回歸分析(Stepwise Regression Analysis)或是訊息增益(Information Gain)的特徵選擇方法,選擇的顯著特徵,但它會失去一些信息。而其他不同的方法,如PCA,ICA等,線性萃取特徵方法,如果問題非常複雜,它們可能沒有什麼用處。然而,特徵選取或特徵萃取的方法對於預測複雜的問題,效果皆不是非常好。因此,本研究將使用核函數主成份分析萃取資訊予交易系統,它是一個非線性映射。根據許多文獻,非線性映射可以解決高維度問題,如股票預測。 故本研究將其應用於股市交易決策系統,模擬數個美國股票的自動交易系統預測連續數年的股票交易,實驗結果顯示,核函數主成份分析的非線性特徵提取方法,相對於ICA與SRA來說,可以獲取更高的利潤,為更加優異的方法,並且預測長期趨勢確實可以獲得優於短期趨勢的利潤。

並列摘要


This paper presented a kernel-based principal component analysis (kernel-PCA) to extract critical features for improve stock trading model performance. The feature extraction method is one of the dimensionality reduction problems (DRP). The kernel-PCA is a feature extraction approach which has been applied to data transformation from known variables to improve capture critical information. The kernel-PCA is a kernel-based data mapping in principal component analysis (PCA) so it has characteristics on PCA and non-linear mapping. Another DRP method, the feature selection method just choice a small set of feature from known variables but among these feature still have collinearity problem so it cannot reflect clear information. However, most feature extraction methods used variable mapping to eliminate the variables noisy and collinearity. In this paper, we use the kernel-PCA method into stock trading model that it will be transform stock technical indices (TI) to produce critical features of smaller dimension. The kernel-PCA method is tested to various stocks and use sliding window testing methods which have half-year and one-year testing strategy. The experimental results show that the proposed method can generate more profits in America stock market than other DRP methods. This stock trading model is very practical for real-world application, and it can be implemented in a real-time environment.

參考文獻


35. 潘依芳,「建構趨勢切割法與支撐向量迴歸於股票買賣時機之預測」,元智大學,碩士論文,2012 年。
15. 賴志銘,「叢集式類神經網路在股價轉折點預測之應用」,元智大學,碩士論文,2008。
17. 張凱婷,「應用支撐向量迴歸及模糊歸責於股價買賣點之預測」,元智大學,碩士論文,2011。
19. 楊雯寧,「臺灣股價指數預測模型之探討」,元智大學,碩士論文,2002 年。
1. Ge, X., “Pattern Matching in Financial Time Series Data,“Final Project Report for ICS 278 UC Irvine ,1998.

延伸閱讀