Selecting a suitable variable combination to predict stock prices using support vector machines

Abstract

With technological development, trading in stock markets has become more accessible to the general public. However, owing to the highly volatile nature of stock prices, stock price predictions remain a challenging task. Literature shows Support Vector Machines as a promising technique. This paper aims at identifying the best variable combination to predict the stock prices using Support Vector Machines along with the application of forward filling and linear interpolation as data filling methods and random search and grid search as hyper parameter optimization methods. After the individual evaluation of all models, data filling method of linear interpolation, hyper parameter optimization method of grid search and independent variable combinations with adjusted close price are found to give better results for prediction of stock prices.