Sentiment Analysis of Crypto Coin on Twitter Data Using Text Mining Method with K-Means Clustering Case Study: Bitcoin, Ethereum, and Binance

Abstract

Since the internet's emergence, numerous facets of daily life have evolved, including online investment opportunities. Cryptocurrency investment has gained popularity in this digital era. Prospective investors now have various cryptocurrencies to choose from, tailored to their preferences. Due to the volatility of cryptocurrency market, investors need to have careful consideration of which cryptocurrency to be chosen including its related decision to be made during their investment. One crucial consideration in this selection process is assessing fellow investors' opinions, often found on social media. This study employs text mining, using the k-means clustering algorithm to explore prevailing sentiments among cryptocurrency investors. The data source consists of Twitter comments on three prominent cryptocurrencies, Bitcoin, Ethereum, and Binance totaling 55,651 tweets. The findings predominantly reveal neutral sentiments towards these cryptocurrencies. Notable positive sentiment topics for Bitcoin include “worth” and “new,” while negative sentiments revolve around “firm” and “bank,” and neutral sentiments are linked to “digital” and “year.” Ethereum exhibits positive sentiments like “good” and “defi,” negative sentiments such as “time” and “long,” and neutral sentiments around “price” and “ethic”. Binance's positive sentiments include “live” and “kind,” a negative sentiment related to “case,” and neutral sentiments encompassing “learn” and “check”. Moreover, the Davies Bouldin Index evaluation for the coin clusters yielded scores of 0.7996 for Bitcoin, 0.7820 for Ethereum, and 0.7149 for Binance. These indices, falling between 0 and 1, signify well-structured clustering.