Stock Market Price Prediction through News Sentiment Analysis & Ensemble Learning

Stock market trends tend to vary according to real-life situations and market sentiments. The growing interest in the Financial-Technology sector has culminated in more research being carried out in this area. For sentiments, news headlines have proven to be a major source and can be used without the further context of the news article. Most of the research for stock market price prediction has been carried out based only on daily stock data or using standard machine learning algorithms. Hence, in this research project, we find the importance of using the sentiment data extracted from news headlines and daily stock data combined for the prediction of trends and stock prices. To carry out market trends prediction we propose the use of the ensemble technique XGBoost, whereas, for price prediction time series, LSTM (Long ShortTerm Memory) cells are used. Our results indicate that combining both data as input using ensemble techniques and time series models gave much better and more accurate output than the standard methods that have been employed before.