On the Construction and Analysis of Time-Series-Oriented Lexicons

• Main Authors: Lai, Cheng-Yi, 賴承毅
• Other Authors: 王釧茹
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/52924983784941283362

碩士 === 臺北市立大學 === 資訊科學系 === 104 === This thesis proposes a novel framework to build a time-series-oriented lexicon which can cover different types of sources and also has explicit links with the targets of prediction problems. In the framework, the input is composed of a text stream, such as financial news and a financial time series, such as the stock prices of a company. We then calculate the Pearson correlation between the frequency series of each word and the stock price series of a company. Although Pearson correlation gives a good idea of how much the two time series are correlated, it has a limitation in capturing the similarity when one of the series is stretched or shifted. To overcome this limitation, we adopt Dynamic time warping (DTW) to handle the problem. Finally, the words with high correlations will be extracted to build the time-series-oriented lexicon. Additionally, we adopt the learned lexicon to construct a model for stock price movement prediction. The experimental results demonstrate that the learned words generally have good prediction ability, which attests the practicability of the proposed idea of capturing a company's keywords via its historical stock prices.