Using Tweets Sentiment Analysis to Predict Stock Market Movement

Abstract

Research shows that news affects stock market movement and indicates the possibility of predicting the market by using the news as a signal to a coming movement with an acceptable accuracy percentage. In this research, we introduce an approach that predict the Standard & Poor’s 500 index movement by using tweets sentiment analysis classifier ensembles and data-mining Standard & Poor’s 500 Index historical data. The data-mining is used to extract the major companies influencing the S&P 500 index, ranking these companies, and finding the market patterns. Sentiment analysis classification is used to determine whether a tweet is positive or negative for a certain company. We show in this thesis that using classifier ensembles such as majority voting classifier formed by Decision Tree, Bernoulli Naive Bayes, leaner SVC classifiers with majority voting selection criteria, and random forest classifier perform better than classic classifiers in classify tweets. Using ensembles classifiers to classifying a number of companies’ news rather than the all 500 leads to a predication model with an accuracy rate above 80%.