A Novel Data Science Approach for Business and Decision Making for Prediction of Stock Market Movement Using Twitter Data and News Sentiments

Abstract

Stock market movement prediction using artificial neural networks is an active field of finance science; many scientists and analysts have shown a strong interest in forecasting stock market behavior. Several methods have been developed for forecasting stock market trend based on the knowledge of historical data. Recent studies have shown that investor emotions such as Twitter and news have a direct influence on stock market behavior. The foremost intention in this research is towards creating a novel methodology using probabilistic neural network (PNN) to forecast stock market movements. First, historical prices are given as an input to the PNN to make predictions about the stock market. Second, investor sentiments such as Twitter and news are collected. Finally, day Twitter sentiment score and day news score are calculated and then fed to PNN as additional inputs with historical prices to evaluate the improvement in accuracy. The developed model is tested on Infosys and Bharti Airtel Limited. Daily stock prices, Twitter data and news data of selected companies were employed to develop this proposed model. For illustrating effectiveness of this methodology, numerical results were clearly presented. For assessing efficiency of this methodology, we have also compared the proposed framework with a few existing frameworks. The results clearly showed that by adding Twitter and news sentiment score as additional inputs with historical data, improved prediction accuracy was observed.