近年來時間序列分析與預測方法一直是很多產業一個重要的研究領域。過去時間序列分析與預測通常透過二種方法:一為統計方法，像是常用的時間序列模型，如穩定型時間序列之ARMA模型，非穩定型時間序列之ARIMA模型，自我相關條件異質變異之GARCH模型，及單根檢定和共整合分析等。另一為機器學習方法，如類神經網路及模糊推論等。評估這些方法之效能大都以預測誤差最小及找出模型參數為主要研究對象。
由於時間序列觀察值資料主要是由幾個重要的潛在因素所構成，如總體經濟相關產業構面指標、主要股市標的物等。若能找出這些個別潛在因素並研究其動態行為，將可改善預測之準確性。因此在本研究中應用了獨立成份分析法來做特徵擷取的資料前處理動作。ICA相較於過去用常用的時間序列模型及機器學習方法建立預測模型，ICA能從混合及有雜訊之時間序列資料裡以動態過程為基礎找出獨立成份信號源，獨立分析分析法不僅在於解開信號之間的相關性(二階統計)，同時也有降低較高階統計相依的目的(四次方拆解)。
本研究以財金資料做為最後之驗證資料，利用二組資料集當作類神經網路的輸入;原始資料集以及經由ICA處理過後的資料集。實證研究發現，經由ICA處理過後之資料當作輸入會比直接利用原始資料當作輸入來的好。本研究亦發現透過ICA建立重要獨立成份之類神經網路模型及利用這些重要獨立成份重建時間序列的類神經網路模型會有較好的預測。

Analysis and prediction of time series data has been a substantial research area of various industries. It is accomplished by two ways; one is statistic approach, including ARMA model for stable time series analysis, ARMA model of unstable time series analysis, Generalized AutoRegressive Conditional Heteroskedasticty model(GARCH), Unit Root Test and Co-integration Analysis, and the other way is machine learning approach such as neural network and fuzzy inference. One way to evaluate the performance of these methods is to minimize the prediction error and to find out the parameters of model.
Owing to the time series observation data is composed of several significant underlying factors such as overall economic indicators and principal stock target etc. the prediction accuracy could be improved if we can identify these individual underlying factors and explore its dynamic behaviors. Therefore ICA is used as a pre-process for extracting features in this study. As compare to other common used time series model and machine learning approach for building prediction models, independent component analysis can chase down independent components in mixed and noisy time series data., it not only loose the correlation among signals, also reduce high level statistical dependence.
The proposed method in this study is evaluated using financial data. Two data sets, original data set and data set with ICA processed are served as inputs of a neural network. Computational results show that model with ICA processed data can perform better than that with original data input, the study also find that building NN prediction model for each factor and recovering the financial time series by applying ICA to these factors will perform the best prediction.

中文部份

林宗永，「證券投資技術分析指標獲利性之實証研究」，碩士論文，國立政治大學企業管理研究所，1995
劉宜峰，「以類神經網路與ARIMA模式預測台灣股市行為之適用性比較」，碩士論文，私立東吳大學會計研究所，pp3-4，1996
葉怡成，「應用類神經網路」，儒林圖書有限公司，1986。
葉怡成，「類神經網路模式應用與實作」，儒林圖書有限公司，1993。

英文部份
Amari, S., Cichocki, A. and Yang, H. A new learning algorithm for blind signal separation, In Advances in Neural Information Processing Systems,Vol.8, pp. 757-763, 1996.
Bartlett, M. S., Movellan, J. R. and Sejnowski, T. J. Face recognition by independent component analysis, IEEE Transactions on Neural Networks, pp.1450-1464, 2002.
Back, A. D. and Weigend, A. S. A First Application of independent Component Analysis to Extracting Structure from Stock Returns, International Journal of Neural Systems, pp.473-484, 1997.
Bell, A. and Sejnowski, T. An information maximization approach to blind separation and blind deconvolution, Neural Computation , Vol.7, pp.1129-1159, 1995.
Comon, P. Independent component analysis-a new concept? Signal Processing, Vol.36, pp.287-314, 1994.
Cardoso, J. Source separation using higher order moments, International Conference on Acoustics, Speech and Signal Processing , pp.2109-2122, 1989.
Cardoso, J. Higher-Order Contrasts for independent component analysis, Neural Computation, Vol.11, pp.157-192, 1999.
Cover, T. M. and Thomas, J. A. Elements of Information Theory, Wiley, 1991.
Cruces, S. and Cichocki, A. Globally Convergent Newton Algorithm For Blind Decorrelation, In Proceedings of the Forth International Workshop on Independent Component Analysis and Blind Signal Separation (ICA 2003), Nara, Japan, 2003.
Douglas, S. On the convergence Behavior of the FastICA Algorithm, In Proceedings of the Forth International Workshop on Independent Component Analysis and Blind Signal Separation (ICA 2003), Nara, Japan , 2003.
Górriz, J. M., Puntonet, C. G., Salmerón, M. and Lang, E. W. Time Series Prediction using ICA Algorithms, IEEE international Workshop on Intelligent Data Acquisition and Advanced Computing Systems: Technology and Applications pp.8-10, Lviv, Ukraine, 2003.
Hamra, S.B., Smaoui, N. and Gabr, M. The Box-Jenkins analysis and neural network prediction and time series modeling, Applied Mathematical Modeling, Vol.27, pp.805-815, 2003.
Hornik, L., Stinchombe, M., and White, H. Multilayer Feedforward Networks Are Universal Approximations, Jorunal of Applie Econometrics, Vol.10, pp347-364, 1989.
Hyvärinen, A. A family of fixed-point algorithm for independent component analysis, In Proc. IEEE Int. Conf. on A caustics, speech and signal Processing, pp.3917-3920, Munich, Germany, 1997.
Hyvärinen, A. and Oja, E. A fast fixed-point algorithm for independent component analysis, Neural computation, Vol.9, No.3, pp.1483-1492, 1997.
Hyvärinen, A. New approximations of differential entropy for independent component analysis and projection pursuit, In Advances in Neural Information Processing Systems, Vol.10, pp.273-279, 1998.
Hyvärinen, A. Fast and Robust Fixed-point Algorithm for independent component analysis, IEEE Transactions on Neural Networks, Vol.10, No.3, pp.626-634, 1999.
Hyvärinen, A. and Oja, E. Independent component analysis: algorithms and applications, Neural Networks, pp.411-430, 2000.
Haykin, S. Neural Networks, Prentice Hall, 1999.
Hild, K., Erdogmus, D. and Principe, J. Blind Source Separation Using Renyi’s Mutual Information, IEEE signal Processing Letters, Vol.8, No.6, pp.174-176, 2001.
Ikdea, S. and Toyama, K. independent component analysis for noisy Data-MEG data analysis, Neural Networks, pp.1063-1074, 2000.
Jutten, C. and Herault, J. Blind separation of source, Part I: An adaptive algorithm based on neuromimetic architecture, Signal Processing , Vol.24, pp.1-10, 1991.
Karhunen, J., Hyvärinen, A., Vigario, R., Hurri, J. and Oja, E. Applications of neural blind separation to signal and image processing, In Proc. Int. Conf. on Acoustics, Speech, and Signal, Munich, Germany, 1997.
Kiviluoto, K. and Oja, E. Independent component analysis for parallel financial time series, In Proc. Int. Conf. on Neural Information Processing, Vol.2, pp.895-898, Tokyo, Japan, 1998.
King, B. F. Market and Industry Factors in Stock Price Behavior, The Journal of Business, Vol.39, No.1, pp139-190, 1966.
Kurita, S., Saruwatari, H., Kajita, S., Takeda, K. and Itakrua, F. Evaluation of blind signal separation method using directivity pattern under reverberant conditions, pp.3140-3143, 2000.
Lee, T. W. Independent Component Analysis; Theory and Applications, Kluwer Academic Publishers, Boston, ISBN: 0-7923-8261-7, pp.7-11; pp.18-21; pp.27-49; pp.83-100, 1998.
Mansour, A. and Kawamoto, M. A Classification of ICA Algorithms According to their Applications & Performances, IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences, Vol.86, No.3, pp.620-633. 2003.
Murata, N., Ikeda, S. and Ziehe, A. An approach to blind source separation on based on temporal structure of speech signals, Neuralcomput, Vol. 41, pp.1-24, 2001.
Miller, E. and Fisher J. ICA using spacings estimates of entropy, In Processings of the Fourth International Workshop on Independent Component Analysis and Blind Signal Separation (ICA 2003), Nara, Japan, 2003.
Moody, J. E. and Wu, L . What is the “true price”?-State space models for high frequency FX data, Process in Neural Information Processing, Springer, Berlin, pp.697-704, 1997.
Niederhoffer,V. The Analysis of World Events and Stock Prices, The Journal of Business, Vol.44, No.2, pp.193-219, 1971.
Park, H. A modification of the gradient algorithm for blind signal separation, In Proceedings of the Forth International Workshop on Independent Component Analysis and Blind Signal Separation (ICA 2003), Nara, Japan, 2003.
Papoulis, A. Probability, Random Variables, and Stochastic Process, McGraw-Hill, 3rd edition, 1991.
Sattar, F., Siyal., M.Y., Wee, L.C. and Yen, L.C. Blind source separation of audio signals using improved ICA method” Proceedings of the 11th IEEE Signal Processing Workshop on Statistical Signal Processing , pp.452-455, 2001.
Tong, L., Liu, R., Soon, V. C. and Huang, Y. Indeterminacy and identifiably of blind identification, IEEE Trans, Circuits, Syst, Vol.39, No.5, pp. 499-509, 1991.
Yang, H. H. and Amari, S. I. Adaptive Online Learning Algorithms for Blind Separation: Maximum Entropy and Minimum Mutual Information, Neural Computation, Vol.9, pp.1457-1482, 1997.