ARIMA

In Rollforward & time series

The ARIMA function is a forecasting function based on the autoregressive integrated moving average (ARIMA) model. Learn more.

Use this function when you want to forecast a time series from historical patterns without defining explicit driver nodes.

Parameters

NodeNode referenceRequired

Node that contains the base data for the forecast. It must contain at least one time level. Predictions are based on the time values in the node and extend to the end of the horizon period.

pNumberRequired

The order (number of time lags) of the autoregressive model.

dNumberRequired

The degree of differencing.

qNumberRequired

The order of the moving-average model.

PNumberRequired

Autoregressive term for the seasonal part.

DNumberRequired

Differencing term for the seasonal part.

QNumberRequired

Moving average term for the seasonal part.

mNumberRequired

The number of periods in each season.

Watch out

Example

Monthly forecast from historical data

This example shows monthly historical passenger data used as the basis for an ARIMA forecast. The result keeps the historical observations and extends the series with forecasted periods to the end of the horizon.

Input node: 'Passenger historic'

Data set: Passenger historic.xlsx

Data horizon: 2000 - 2011

Formula: ARIMA('Passenger historic', 1, 1, 1, 1, 0, 1, 12)

ARIMA horizon: 2000 - 2020

The rows through 2011 repeat the historical input values unchanged; the rows from 2012 onward are the new forecasted periods.

Chart:

See also

ROLLFORWARD_ADVANCED

When you want to project values with explicit driver nodes instead of forecasting from historical patterns alone.

MOVINGAVG

When you want to smooth historical values over a rolling window instead of creating a forecast.