Time-Series Forecasting of Ethereum Price Using Long Short-Term Memory (LSTM) Networks

Abstract

The cryptocurrency, ether (ETH), often sees its price go through rapid fluctuations due to its popularity for being the foundation for various decentralized applications and the fuel of the Ethereum network which has simplified commerce and trade between both anonymous and recognized parties, unfavourable circumstances like political conflicts, natural disasters, and so on, causing the market to become extremely volatile and risky for the crypto investors and the developers. So, from the urge to have a specialized ether price forecasting system, this research aimed to find an accurate price prediction model for ether using the long short-term memory (LSTM) network. For this, ether time-series price data were fitted into multiple basic and hybrid LSTM network variants, with future prices predicted using both univariate and multivariate time-series analysis. Furthermore, a comparative analysis was conducted among the models and also some popular existing forecasting techniques like autoregressive integrated moving average (ARIMA) as the baseline forecast to understand the effectiveness of the LSTM networks, especially, the hybrid variant, in the prediction of future market behaviour.