Sentiment Analysis on Cryptocurrency using Youtube Comments

Abstract

Because of the rising popularity of cryptocurrency in the world, it is essential in these times to understand the market sentiment to make predictions of price and make investment related decisions. Therefore, a model is designed to classify YouTube comments based on cryptocurrency. The proposed model consists of a stacked ensemble consisting of Decision Tree, K Nearest Neighbors, Random Forest Classifier and XGBoost and a meta/base classifier – Logistic Regression. The proposed model achieves an accuracy of 94.2%. In addition, based on our research, we've come to several important findings and takeaways about the current state of cryptocurrencies around the world.