Investigating the Role of Reddit in Cryptocurrency Price Fluctuations

Abstract

In recent years, the exponential growth in the value of cryptocurrencies like Bitcoin and Ethereum has attracted many investors. Unlike traditional currencies, cryptocurrencies do not have a central issuing or regulating authority; instead, they use a decentralized system to record transactions and issue new units. This leads to their prices being influenced by many factors, including their popularity on social media platforms. Consequently, sentiment extracted from these platforms can potentially act as an indicator of changes in cryptocurrency prices and help in their prediction. This thesis focuses on discussions on Reddit, a popular social media site, and aims to answer the research question: “Can sentiment features extracted from Reddit discussions help predict the volatile price movements of cryptocurrencies?” To address this question, discussions about Bitcoin and Ethereum on Reddit were extracted and analyzed using two sentiment analyzers: VADER and CryptoBERT. The sentiment and textual features extracted from these discussions were used to create time series data, which were then utilized for forecasting cryptocurrency prices. Both statistical models, such as Vector Autoregression (VAR), and advanced machine learning models, such as Long Short-Term Memory (LSTM) networks, were employed for the prediction task. The findings of this study illustrate that the inclusion of sentiment features only from Reddit discussions does not significantly help in the prediction of cryptocurrency prices. This outcome suggests that while discussions on Reddit reflect varied public opinions, they do not provide sufficient information for predicting the price movements of cryptocurrencies. This could be attributed to cryptocurrency prices being influenced by many factors, which may not be adequately represented by sentiment data alone. Moreover, the inherent volatility of cryptocurrency prices further adds to the complexity of the task. These fluctuations in prices can often overshadow the sentiment trends extracted from social media platforms.