Forecasting Daily Ethereum Closing Price: An Autoregressive Integrated Moving Average (ARIMA) Approach

Abstract

Ethereum, a leading digital asset by market value, has gained increasing attention from investors and researchers because of its high price volatility and market unpredictability. This study forecasts Ethereum cryptocurrency daily closing prices using the Box-Jenkins Autoregressive Integrated Moving Average (ARIMA) methodology, drawing on data from January 1, 2019, to December 31, 2025. Stationarity analysis via the Augmented Dickey-Fuller ADF and Kwiatkowski–Phillips–Schmidt–Shin (KPSS) tests confirmed that first differencing was required to render the series suitable for the modeling. Through systematic model identification, estimation, and comparison of ten candidate ARIMA specifications, the ARIMA(1,1,0) model emerged as the optimal fit, yielding the lowest information criterion values of Akaike information criterion, Bayesian information criterion (AIC = 24,943.883; AICc = 24,943.84; BIC = 24,955.12). Residual diagnostic tests, including the Ljung-Box test for serial correlation, the Autoregressive Conditional Heteroskedasticity (ARCH-LM) test for heteroscedasticity, and the Shapiro-Wilk test for normality, confirmed that the model residuals are free of serial dependence, although they exhibit time-varying volatility and non-normal distribution, features commonly associated with financial time series. The fitted model was subsequently applied to generate 30-day ahead forecasts with 95% confidence intervals, revealing relatively stable price expectations in the near term alongside progressively widening prediction bands that reflect growing uncertainty over longer horizons. These findings underscore the practical utility of the parsimonious ARIMA(1,1,0) model as a transparent and accessible tool for short-term Ethereum-price forecasting and investment risk assessment.