Machine Learning for Predicting Bitcoin and Ethereum Price Fluctuations with News Analysis

Abstract

Cryptocurrencies like Ethereum and Bitcoin are highly volatile, offering both advantages and disadvantages to financiers and traders. Accurately forecasting price variations is crucial in such a dynamic environment. The intricacies of cryptocurrency trading, including the impact of emotions, often surpass conventional methods. Incorporating media scrutiny, our research proposes an improved collective system for forecasting Bitcoin and Ethereum price changes. Our integrated model combines the strengths of multiple models, leveraging historical price data and sentiment analysis of news articles to capture the influence of news sentiment on price movements. We gather news from credible sources and collect historical price data from cryptocurrency trading platforms to construct our database. The ensemble model generates more reliable predictions by combining predictions from multiple models, reducing volatility and biases. Experimental results show that our model outperforms baseline approaches, accurately predicting price swings for Bitcoin and Ethereum. Incorporating news analysis significantly improves predictive accuracy, emphasizing the importance of considering external variables in forecasting Bitcoin prices.