Performance Evaluation of Time Series Forecasting Methods in The Stock Market: A Comparative Study

Stock market forecasting is a challenging area for many researchers and investors since it is a stochastic environment. This paper aims to apply some of the existing forecasting methods to find the optimal method that gives high accuracy based on the given data. IBM stock market dataset has been used for this paper. Ten years of numerical collected data have been applied for training and testing. Eight forecasting methods have been tested which are Linear Regression, Multilayer Perceptron, RBF Regressor, SMO reg, Bagging, Random Sub Space, Timeseries Holt-Winters, and Random Forest. Python and Weka tools have been used for processing, analyzing, and testing. SMO regression has shown an excellent performance compared to the other applied methods in IBM stock market forecasting. The outcome of this paper will help the investors to take the right decision for selling, holding, or buying in the stock market.