A Complete VADER-Based Sentiment Analysis of Bitcoin (BTC) Tweets during the Era of COVID-19

Toni Pano,Rasha Kashef

doi:10.3390/bdcc4040033

Abstract

During the COVID-19 pandemic, many research studies have been conducted to examine the impact of the outbreak on the financial sector, especially on cryptocurrencies. Social media, such as Twitter, plays a significant role as a meaningful indicator in forecasting the Bitcoin (BTC) prices. However, there is a research gap in determining the optimal preprocessing strategy in BTC tweets to develop an accurate machine learning prediction model for bitcoin prices. This paper develops different text preprocessing strategies for correlating the sentiment scores of Twitter text with Bitcoin prices during the COVID-19 pandemic. We explore the effect of different preprocessing functions, features, and time lengths of data on the correlation results. Out of 13 strategies, we discover that splitting sentences, removing Twitter-specific tags, or their combination generally improve the correlation of sentiment scores and volume polarity scores with Bitcoin prices. The prices only correlate well with sentiment scores over shorter timespans. Selecting the optimum preprocessing strategy would prompt machine learning prediction models to achieve better accuracy as compared to the actual prices.

Highlights

Recent research studies have emerged that involve the impact of COVID-19 on the financial market, including cryptocurrencies [1,2,3,4,5,6,7,8]
The purpose of this paper is to perform a complete Valence Aware Dictionary and sEntiment Reasoner (VADER)-based sentiment analysis of BTC tweets during the era of COVID-19 to identify the role of different preprocessing strategies in predicting Bitcoin prices
We identified the optimal preprocessing strategy of Bitcoin tweets introduced in the VADER-based Sentiment Analysis during the era of COVID-19

Summary

Introduction

Recent research studies have emerged that involve the impact of COVID-19 on the financial market, including cryptocurrencies [1,2,3,4,5,6,7,8]. Symeonidis et al compared the significance of different preprocessing techniques for sentiment analysis of tweets [10] They used four different machine learning algorithms, Linear Regression, Bernoulli Naïve Bayes, Linear Support Vector Machine, and a Convolutional Neural Network to classify tweets as positive, negative, or neutral sentiment. The purpose of this paper is to perform a complete Valence Aware Dictionary and sEntiment Reasoner (VADER)-based sentiment analysis of BTC tweets during the era of COVID-19 to identify the role of different preprocessing strategies in predicting Bitcoin prices.

Objectives

Findings

Discussion

Conclusion