Learning to trade on sentiment

The increasing availability of big data has made it possible to research the sentiment influence to the individual company. We use investment social media data to extract the sentiment expressed in the financial news articles by applying deep learning model, Long Short-Term Memory (LSTM) neural network. The textual sentiment (bullish or bearish idea) can be classified by all the machine learning classifiers and deep learning models and even some traditional dictionary approaches. Based on our experiments, we have found that the Long Short-Term Memory (LSTM) neural network performs best with the accuracy at 94%. Based on the sentiment related with individual company, we build a market-neutral trading strategy called majority votes strategy to perform a comprehensive study on how the sentiment of the individual company influence the financial returns. In this paper, we demonstrate how financial sentiment analysis can be utilized to build trading strategy by incorporating the sentiment factor.