Turning tweets into trades
The thesis completed for my degree was titled 'Turning Tweets Into Trades - Sentiment Analysis for Directional Stock Price Predictions'. It was an individual research project involving three areas of study: machine learning, sentiment analysis, and automated trading.
A novel approach was used to build an algorithm which learnt from stock-related tweets, and could be used to make predictions on new and unseen tweets. These predictions were input to an automated trading system, and over a 1 month testing period 468 trades were placed. The algorithm performed significantly better than random chance and outperformed two baseline sentiment analysis approaches tested, generating a monthly return of 5%.