Sentiment Analysis and Cryptocurrency Price Correlation: A Data-Driven Study

Abstract

<ns3:p>The increasing integration of Artificial Intelligence (AI) and Natural Language Processing (NLP) in financial markets has revolutionized the predictive modeling of asset prices. In cryptocurrency markets, where price movements are largely driven by investor sentiment, sentiment analysis has emerged as a valuable tool for understanding market behavior. This study investigates the correlation between sentiment polarity extracted from FinBERT and FinancialBERT—two pre-trained NLP models optimized for financial text analysis—and the price fluctuations of Bitcoin (BTC), Ethereum (ETH), and Ripple (XRP). The research explores the role of sentiment indicators as leading signals for price trends by examining their correlation across different time lags (immediate, 12-hour, and 24-hour periods).The study utilizes a hybrid sentiment model which uses FinBert and FinancialBert using time lagged correlation models, aggregating sentiment scores from multiple financial news sources retrieved via the MediaStack API, while historical cryptocurrency prices were obtained from the CoinGecko API. A dataset of 1,300 news articles over 90 days was analyzed, revealing that Ethereum exhibited the strongest sentiment-price correlation (0.3819, increasing to 0.3900 after 24 hours), followed by Bitcoin (0.2899 to 0.2919) and XRP (0.1005 to 0.1205). These findings suggest that market sentiment has a delayed impact on price movements, with Ethereum being the most responsive to sentiment fluctuations. This research highlights the potential of AI-driven sentiment analysis as a supplement to traditional financial indicators, offering new opportunities for algorithmic trading and risk assessment in decentralized finance (DeFi) markets. Future research should explore real-time applications, multilingual sentiment tracking, and hybrid predictive models to enhance the accuracy of sentiment-based cryptocurrency forecasting.</ns3:p>