Forecast a time series using the Naive2 method

Produces forecasts using the Naive2 benchmark from the M4 forecasting competition. The function first applies test_seasonality() at the specified frequency. If seasonality is not detected, the final observation is repeated for the complete forecast horizon.

If seasonality is detected, the series is adjusted using classical multiplicative decomposition. A naive forecast is then produced from the seasonally adjusted series and reseasonalized using the estimated seasonal factors.

Usage

forecast_naive2(x, freq, n_ahead)

Arguments

x: A numeric vector containing the observed time series.
freq: Integer value. A positive whole number specifying the seasonal frequency, such as `12` for monthly data or `4` for quarterly data.
n_ahead: Integer value. A positive whole number specifying the number of future observations to forecast.

Value

A numeric vector of length `n_ahead` containing the point forecasts.

Examples

x <- as.numeric(AirPassengers)

forecast_naive2(
  x = x,
  freq = 12,
  n_ahead = 18
)
#>  [1] 437.4820 424.6949 484.1683 469.0476 471.6776 534.8310 589.5167 586.3231
#>  [9] 509.7019 443.0222 385.0685 432.0000 437.4820 424.6949 484.1683 469.0476
#> [17] 471.6776 534.8310