BITCOIN PRICE FLUCTUATIONS AND GOOGLE NEWS WITH MACHINE LEARNING TECHNIQUES

Abstract

This research investigates the predictive power of news sentiment from Google News on Bitcoin price movements, leveraging a five-year dataset of news headlines (2019 to 2024). By correlating sentiment scores with historical Bitcoin prices, the study employs various machine learning algorithms to forecast price trends. The results indicate that while Decision Tree and Random Forest models offer balanced predictions, Logistic Regression and Support Vector Machines achieve high AUC scores but suffer from class imbalance. In contrast, Naïve Bayes and KNN models prove less effective. The findings suggest that sentiment analysis of news headlines can provide moderate short-term predictions for Bitcoin price fluctuations. This study introduces an innovative tool for investors and market analysts, offering insights into the influence of news sentiment on cryptocurrency prices.