Analysing crypto news sentiment to predict bitcoin prices

Abstract

The purpose of this paper is to analyse the ability of news content to predict cryptocurrency markets. The role of news announcement is central to pricing and revisions in pricing of any asset. Right from the 16th century news of ships arriving at ports with tradable goods resulted in the fluctuation of local market prices. In the digital age where TV screen flashes breaking news at microsecond frequency, the influence on the prices has never been more profound. Using news articles to predict markets has always been a bottleneck for traders, the major issue being transforming words and semantics to financial numbers. The meteoric rise of natural language processing in other fields has finally made this task possible for humans. Some common ways include transforming raw texts to bags of words, one hot encoding or advanced word embeddings and feeding them to ML models which is attempted in this paper.