标题: 应用技术指标于台湾股票市场风险系数Beta值之研究
Application of Technical Indicators for Beta Coefficient in Taiwan Stock Market
作者: 林秋红
Lin, Chiou-Hung
陈安斌
Chen, An-Pin
资讯管理研究所
关键字: beta系数;技术指标;KNN演算法;台湾股票市场;β Coefficient;Technical Indicator;k-Nearest Neighbors;Taiwan Stock Market
公开日期: 2015
摘要: 由于近期欧美股市掀起了关于“Smart Beta”的讨论热潮,并愈来愈受全球投资者对β值的关注。因此,本研究结合大数据分析与财金的领域,为使投资人降低风险,运用预测股票市场的风险系数β值作为股市风险衡量的依据。
本研究以五个技术指标乖离率(BIAS)、相对强弱指标(RSI)、威廉指数(Williams %R)、动向指数(DMI)、心理线(PSY)等作为投资标的特征值,利用滚动式财务预测,并结合KNN演算法,从过去的巨量历史资料中找出短(21天)、中(65天)、长(250天)期之五个特征值最相似的β值样本,作为未来β值的预测基础,透过这些技术指标背后所隐含的物理力量,尝试找出更贴近实务面的β值,再与实际β值加以比对,并以台湾股票市场中重要的权值股台积电及TW50为例。
研究结果显示与大盘连动性高的TW50以中、短天期为基础之β预测值较为准确,而台积电则以短天期的β预测值较为准确。希望透过本研究能够让投资者在进入股市之前,能更精准地掌握个股的风险,做出最佳的投资策略,以便投资人能够降低风险并获得最大的利润。
Due to the recent lively discussion on "Smart Beta" in Europe and the US stock market, it has begun growing interest in β value by global investors. Therefore, this research paper combined two fields of knowledge, big data analysis and finance. In order to reduce the risk for investors, we applied β coefficient of stock market to measure the risk of a stock.
Five technical indicators such as Bias Ratio(BIAS), Relative Strength Index(RSI), Williams Overbought / Oversold Index (Williams %R), Directional Movement Index(DMI), and Psychological Line(PSY) are used as characteristic values of an investment stock target. Through physical forces behind these technical indicators, we apply KNN algorithm from past historical data to identify each of the different days of the five most similar characteristic values to predict the most similar β samples to find out more practical β value, and then compare with the actual β value. Studies have shown that the short-day period and medium-day period basis is more accurate predictive value of β for TW50, and short-day period basis is more accurate predictive value of β for TSMC.
In this study, we expect that before the investors entering the stock market can properly grasp the risk of an individual stock, and make the best profits from stock-picking strategy.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT079834805
http://hdl.handle.net/11536/126695
显示于类别:Thesis