Sentiment Analysis for Bitcoin Price Prediction via Machine Learning

Abstract

Cryptocurrencies in the past years have gained more and more popularity, starting to be known and traded increasingly. In particular, among all, Bitcoin can be regarded as the most widespread, and, consequently, the most discussed, not only in real life but also, and especially, on the Internet on social media. This is the reason why, along with Twitter data, it constitutes the foundation of this dissertation, whose aim is to perform Sentiment Analysis in order to study the relationship between tweets on Bitcoin and its market price variation over a three-year time horizon.

Specifically, we want to verify if the so-called vox populi possesses a sort of predictive power that allows us to improve Bitcoin closing price prediction. A step-by-step procedure will be adopted, increasing more and more the complexity both in terms of models applied and data considered, the last step being the application of a Multilayer Perceptron Artificial Neural Network.