B3 Stock Price Prediction Using LSTM Neural Networks and Sentiment Analysis

Abstract

This article presents an approach to predict stock prices which incorporate sentiment analysis from Twitter posts as an input to an Long Short Term Memory (LSTM) Neural Network to help in the decision process. The sentiment analysis measures subjectivity and polarity as well as the number of tweets about the company to capture the market mood, which influences the stock prices, were evaluated. The main company used to evaluate our method is Vale (VALE3). The sentiment analysis helps to reach an Root Mean Squared Error (RMSE) of 0.021. We also validate our method with JHSF (JHSF3) and Usiminas (USIM3), obtaining RMSE of 0.012 and 0.016, respectively.