Prediction of Bitcoin prices using Twitter Data and Natural Language Processing

Abstract

The influence of social media platforms like Twitter had long been perceived as a bellwether of Bitcoin Prices. This paper aims to investigate if the tweets can be modeled using two different approaches, namely, the Naïve Bayes and LSTM models, to compute the sentiment scores in order to predict the Bitcoin price signal. Through the experiments conducted, the LSTM model indicates some degree of predictive advantage compared to the Naïve Bayes model.