Prediction of the Moving Direction of Google Inc. Stock Price Using Support Vector Classification and Regression

Abstract

Forecasting the short-term trend of a stock market has long been a big challenging task.   Parameters of stock markets, including open/close prices, daily-high/low prices and trading volumes, were frequently used in previous studies to forecast the stock market. Basing on the fact that the moving direction of these parameters have certain inertia within short-term period, we here explored the potential application of the moving trends of these parameters within 4 different time periods (5, 15, 30 and 45 trading days respectively) for forecasting the movement direction of stock price of Google Inc. by using support vector classification (SVC) and support vector regression (SVR).  We found that among the 4 different time periods tested, the moving trend within 30 days has the best accuracy on the prediction of the stock price of Google Inc., and using SVC and SVR combination improved the prediction performance. These results indicated that moving trends of stock transaction data within a certain time period have good inertia and are thus useful for forecasting the moving direction of stock price.