Forecasting Prices of Financial Assets with Support Vector Regression and Multiple Liner Regression: Evidence from both Stock Indexes and Individual Stocks

• Main Authors: LIU,YU-CHUN, 劉昱均
• Other Authors: HSU,CHIH-MING
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/4cespu

碩士 === 明新科技大學 === 管理研究所碩士班 === 104 === Due to the booming development in the financial market of Taiwan, financial product has become more diversified, meanwhile, new innovative derivative financial product based on Taiwan weighted stock index and OTC (over-the-counter) stock index as target emerges to the market too. The importance of these two indexes to the stock market is obvious. TSMC (2330) is the largest company in Taiwan with stock listed in the stock market, meanwhile, it is also the favorite investment target for foreign investor and the general public. In recent years, with the backup of the trends such as financial technology, internet of things and e-commerce, PChome Online Inc. (8044) in the OTC market has become the star of the stock market. Therefore, in this study, support vector regression model will be used as basis to be associated with several technical indexes such as KD value, RSI and MACD to construct three prediction models to predict respectively stock average (Taiwan weighted stock index and OTC stock index) and the highest price, lowest price and closing price of individual stock (TSMC, PChome Online Inc.), meanwhile, the results will be compared with multiple regression model for the related prediction capability. In this study, daily data were used for empirical study, and the data taken period was from Jan. 01, 2010 to June 30, 2015, wherein the period from Jan. 01, 2010 to Dec. 31, 2012 was training period, period from Jan. 01, 2013 to June 30, 2015 was prediction period. Meanwhile, mean squared error (MSE) and mean absolute percentage error (MAPE) were used in this study to make comparison on the prediction capability of the empirical model. The empirical result shows that in predicting the closing price of Taiwan weighted stock index using multiple regression in association with model 3 (high opening and low closing in association with technical index), the prediction result will be the best, which were similarly true in predicting the closing price of TSMC and in predicting OTC index and the closing price of PChome Online Inc. This empirical result is helpful for the investor to seize the stock average and the daily stock price information and fluctuation of individual stock, consequently, the investor’s investment performance can be enhanced.