Abstract :
Abstract Cryptocurrencies, functioning as digital currencies, undergo regular fluctuations in the present market, reflecting the emotional aspect of the cryptocurrency realm. It is a well-established fact that sentiment is linked to Bitcoin and Ethereum values, employing a Twitter-based strategy to predict changes. While prospective Bitcoin returns do not display a correlation with emotional variables, indicators of emotions tend to anticipate Bitcoin exchange volume and return volatility. Emotions wield an influence over a broad spectrum of financial investor returns, thereby, potentially affecting market dynamics by triggering significant price shifts. The research delves into gauging emotional factors extracted from 2,050,202 posts on Bitcointalk.org, investigating how these emotions impact Bitcoin's price fluctuations. We have used a unified dataset named 'data F' in which all categories of emotions are consolidated. Subsequently, data preprocessing steps are implemented to cleanse the dataset. Two feature engineering techniques, namely TF-IDF and BoW are employed. The research explores ten supervised machine learning (ML) models as classifiers, with four of these models (LR, Stochastic Gradient Descent, SVM and GB) yielding the highest accuracy at 0.93%.
Keyword :
N/A