A Research of Stock Price Index in Taiwan Based on Text Sentiment Analysis

• Main Authors: Yi-Chi Chou, 周以騏
• Other Authors: Sung-Shun Weng
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/u6y9u4

碩士 === 國立臺北科技大學 === 資訊與財金管理系 === 106 === Related issues of stock investment have always been quite concerned whether investors, investment experts or other people still hesitating to get into the stock market. Although we can speculate on the future trend from the historical pulsation of the stock market, relevant news still have some effects on the short-term stock market, but it has always been easily ignored by the public. With the development and popularization of community forum media, which has now become a tool most people use to browse and discuss. However, investors are not able to pay attention to the social media article information every day and turn it into an investment that can be used to make related decisions for investment. Therefore, based on the general technical index, this study uses different sentiment indicators to generate different forecasting model and input variable combinations in predicting the growth rate of the stock price index, including "technical index", "technical index + agreement index ", "technical index + bullishness index" and "technical index + agreement index + bullishness index ". Finally, this study finds out the cost and gamma parameters for each input variable combination by the support vector regression model. After training and testing for the model, we found that adding two different sentiment index into the technical index can effectively reduce the error rate of prediction for the stock price growth rate. It is the best combination of model variables in this study.