Forecasting EUR/USD with ARIMA and SARIMA

Summary

The document explains univariate time-series forecasting with ARIMA, whose autoregressive, differencing, and moving-average terms are set by the orders p, d, and q. It describes choosing these terms with PACF and autocorrelation plots and checking stationarity with the Augmented Dickey-Fuller test. A worked example uses daily EUR/USD data obtained through MetaTrader 5, then discusses fitting the model, making out-of-sample forecasts, reviewing residual plots, and extending the approach to seasonal ARIMA.

The example provides a stationarity test result for the price series and discusses diagnostic plots and forecast evaluation. The approach relies on past values of one series, so it does not incorporate other market drivers. The article also notes that parameter selection matters and that ARIMA is intended for non-white-noise series; its example does not establish that the forecasts produce a profitable trading strategy.

Key ideas

• ARIMA combines lagged observations, differencing, and lagged forecast errors to model a time series.
• The p, d, and q orders represent autoregressive lags, differencing, and moving-average error lags.
• PACF and autocorrelation plots can help guide order selection and assess differencing.
• The EUR/USD example checks stationarity with the Augmented Dickey-Fuller test before forecasting.
• Residual diagnostics and out-of-sample evaluation are part of assessing an ARIMA fit.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.