預測股價變化是一項具挑戰性的任務。本研究整合支援向量迴歸(support vector regression, SVR)與多元適應性雲形迴歸(multivariate adaptive regression splines, MARS)建構股價預測模式。SVR為近年來廣被重視與應用於各領域的預測方法之一,然而其卻不具有變數篩選的能力。由於在建構股價預測模式時,預測變數的選擇通常會影響模式的績效,因此研究使用處理速度快且具有選擇變數能力MARS做為SVR的前處理工具,提出一個有效的股價指數預測模式。本研究以台灣加權股價指數為實證對象。實證結果顯示,所提之整合MARS與SVR模式無論是在預測誤差或是趨勢預測準確度的表現上均較多元線性迴歸(MLR)、直接使用MARS及直接使用SVR模式為佳。並且在整合模式中,透過MARS能夠從眾多預測變數中篩選出重要的預測變數,藉由分析變數選擇的結果,可對所建構的預測模式有更深入的解讀,提供投資者更多的參考資訊。
In anfinancial company, stock index forecasting is one of the most important and challenging tasks. In this study, wecombined multivariate adaptive regression splines (MARS) and support vector regression (SVR) for stock price forecasting. The proposed model first uses the MARS to select important forecasting variables. The obtained significant variables are then served as the inputs for building the SVR forecasting model. Experimental results from TAIEX revealed that the obtained important variables from MARS can improve the forecasting performance of the SVR models. The proposed combined model outperforms the results of using single multivariate linear regression (MLR), single SVR and single MARS models and hence provides an efficient alternative for stock index forecasting.