財務時間序列的未來走勢一直是投資人所關心的議題,因此,投資人常以過去的財務時間序列資料為對象進行分析,以建立具有良好預測能力的模式,然而,財務時間序列資料中的雜訊常會干擾模式的預測能力。因此,本研究以空間性及時間性獨立成份分析方法處理雜訊問題,再以分類迴歸樹建立投資決策模式,提供投資人客觀的投資決策。為了驗證本模式的穩健性及有效性,本研究以美國道瓊工業指數、香港恆生指數以及臺灣加權指數為對象進行實證分析,研究結果發現,本研究所提模式除了具有良好的預測能力外,績效亦優於傳統方法。
Financial time series forecasting is an important issue for the investors. However, the accuracy of the financial time series forecasting model is always affected by the noise. In order to properly handle this problem, spatiotemporal independent component analysis is adopted in this research. Besides, classification and regression tree is also adopted to build a financial decision making model to provide the investors with objective investment suggestions. Finally, Dow Jones Industrial Average Index, Hang Seng Index, and Taiwan Stock Exchange Capitalization Weighted Stock Index are used to verify the efficiency and robustness of our proposed model. The results show that our proposed model outperforms other models.