使用具代表性之特徵預測股市趨勢

Abstract

為了保護投資大眾以及維持投資市場的健全，證券交易法相關法令規定上市公司必須定期地揭露財務報表，這些財務報表不但可以幫助投資決策，還有助於股票市場的研究與分析。在本篇論文中，我們提出一個有效率的分群方法叫「HRK」，用來預測財務報表發布後的短期股價趨勢。首先，我們將每一篇財務報表轉換成一個特徵向量。我們提出的方法主要包括三個階段。第一階段，利用階層式聚合分群演算法將所有的特徵向量分成數個群集。第二階段，我們遞迴地利用K-means演算法將上一階段的每個群集再分成數個群集，直到每個群集中大部份的特徵向量都屬於同一個類別。然後，我們以質心來代表得到的每個群集，這些質心即為具代表性之特徵向量。第三階段，我們用這些具代表性之特徵向量來預測股價變動的趨勢。實驗結果顯示，不論是預測的準確率或平均獲利，我們所提出的方法皆優於支援向量機方法。

With disclosure regulation, a large amount of financial reports is available for investment purposes and research analysis. In this thesis, we propose an effective clustering method, HRK (Hierarchical agglomerative and Recursive K-means clustering), to predict the short-term stock price movements after the release of financial reports, where each financial report is represented by a feature vector. The proposed method consists of three phases. First, we use the hierarchical agglomerative clustering (HAC) algorithm to divide the feature vectors into several clusters. Second, for each cluster, we recursively divide the feature vectors within the cluster into several clusters by the K-means algorithm until most feature vectors in each cluster have the same label. Then, we compute the centroid for each cluster. The centroids are called the representative feature vectors of the clusters. Finally, we use these representative feature vectors to predict the stock price movements. The experimental results show that the proposed method outperforms the SVM method in terms of accuracy and average profits.