Predicting price changes in ethereum

Neha Narwal, Mila Schultz, Matthew Chen

Abstract: The volatility and price predictability of cryptocurrencies have been a subject of extensive research, primarily focusing on Bitcoin. However, Ethereum, the second-largest cryptocurrency by market capitalization, possesses unique functionalities that may influence its price behavior differently. This paper investigates the predictability of Ethereum price changes using various machine learning models, leveraging historical price data sampled at hourly intervals. We employ a range of classification techniques, including Logistic Regression, Support Vector Machines, Naïve Bayes, Random Forest, and neural networks, as well as time-series forecasting models such as Auto Regressive Integrated Moving Average (ARIMA). Our results demonstrate that the ARIMA model outperforms other approaches, achieving the highest prediction accuracy. The findings highlight the challenges associated with predicting Ethereum price movements and underscore the effectiveness of time-series-specific models in capturing its complex dynamics. These insights contribute to a deeper understanding of cryptocurrency market behavior and inform future predictive modeling efforts.