利用技術指標與總體經濟變數預測台灣股市股價漲跌之研究 : 以支持向量機建構

Abstract

近十年來，SVM(Support Vector Machines)被廣泛運用於資料分類及迴歸處理，支持向量機由於其在許多領域良好的表現而受到矚目。支持向量機中包含一個學習演算法和一個輸入空間，輸入空間內含一個訓練集和一個測試集。藉由訓練集的輸入，學習演算法可以找出一個辨別器，並經由測試集我們可以知道此辨別器的辨別正確度。支持向量機的目的是找出正確度足夠的分類器，針對日後未知的輸入作辨別。基本分析與技術分析是股市投資者使用最廣的分析方法，藉由分析的結果可幫助投資者作出較正確的買賣決策，進而在股市中獲利。本研究即利用SVM之迴歸特性，試圖在基本分析與技術分析的理論基礎下，找出台灣股市未來可能之走勢。在SVM架構下，透過此兩種指標的預測模式分析，可發現台灣股市並非無法預測，在預測隔月股價漲跌方面，所預測的電子類個股，大多有不錯的準確率。整體而言，股本較小的個股使用技術分析訓練的模型有較佳的準確率，股本較大的個股，兩種方法皆有不錯的準確率(45.5%~69.4%)，但就股本較大的個股而言，基本分析模式的表現較佳，部分個股更高達60%以上。

Support vector machine (SVM) is used to classify data in the input space. SVMs are applied to many fields, and have good performance in a wide variety of applications. In this paper, we first introduce the support vector machines. The basic analysis and technique analysis both are the methods that often be used by most of the investors, by analytical of the result can help the investors to make the righter decision, then making a profit in stock market. In recent years, SVM extensive be used in the data classification, and regression. This research used the regression characteristic of the SVM, trying to find out the possible trend in the Taiwan stock market. Through the analytical result found that the Taiwan stock market is can to predict in the aspects of predicting the stock price next month, the electronics stocks chosen have good accurate rate mostly. By all accounts, there have a better accurate rate for the small capital stocks that use the technique analysis, for big capital stocks, two kinds of methods all have good accurate rate (45.5%~69.4%), but the technique analysis is better for big capital stocks, there are a few stocks being up to 60% above.