Feed-forward neural networks with a single hidden layer and lagged inputs for forecasting univariate time series.

nnetar(y, p, P = 1, size, repeats = 20, xreg = NULL, lambda = NULL,
  model = NULL, subset = NULL, scale.inputs = TRUE, x = y, ...)



A numeric vector or time series of class ts.


Embedding dimension for non-seasonal time series. Number of non-seasonal lags used as inputs. For non-seasonal time series, the default is the optimal number of lags (according to the AIC) for a linear AR(p) model. For seasonal time series, the same method is used but applied to seasonally adjusted data (from an stl decomposition).


Number of seasonal lags used as inputs.


Number of nodes in the hidden layer. Default is half of the number of input nodes (including external regressors, if given) plus 1.


Number of networks to fit with different random starting weights. These are then averaged when producing forecasts.


Optionally, a vector or matrix of external regressors, which must have the same number of rows as y. Must be numeric.


Box-Cox transformation parameter.


Output from a previous call to nnetar. If model is passed, this same model is fitted to y without re-estimating any parameters.


Optional vector specifying a subset of observations to be used in the fit. Can be an integer index vector or a logical vector the same length as y. All observations are used by default.


If TRUE, inputs are scaled by subtracting the column means and dividing by their respective standard deviations. If lambda is not NULL, scaling is applied after Box-Cox transformation.


Deprecated. Included for backwards compatibility.

Other arguments passed to nnet for nnetar.


Returns an object of class "nnetar".

The function summary is used to obtain and print a summary of the results.

The generic accessor functions fitted.values and residuals extract useful features of the value returned by nnetar.


A list containing information about the fitted model


The name of the forecasting method as a character string


The original time series.


The external regressors used in fitting (if given).


Residuals from the fitted model. That is x minus fitted values.


Fitted values (one-step forecasts)


Other arguments


A feed-forward neural network is fitted with lagged values of y as inputs and a single hidden layer with size nodes. The inputs are for lags 1 to p, and lags m to mP where m=frequency(y). If xreg is provided, its columns are also used as inputs. If there are missing values in y or xreg, the corresponding rows (and any others which depend on them as lags) are omitted from the fit. A total of repeats networks are fitted, each with random starting weights. These are then averaged when computing forecasts. The network is trained for one-step forecasting. Multi-step forecasts are computed recursively.

For non-seasonal data, the fitted model is denoted as an NNAR(p,k) model, where k is the number of hidden nodes. This is analogous to an AR(p) model but with nonlinear functions. For seasonal data, the fitted model is called an NNAR(p,P,k)[m] model, which is analogous to an ARIMA(p,0,0)(P,0,0)[m] model but with nonlinear functions.


fit <- nnetar(lynx) fcast <- forecast(fit) plot(fcast)
## Arguments can be passed to nnet() fit <- nnetar(lynx, decay=0.5, maxit=150) plot(forecast(fit))
## Fit model to first 100 years of lynx data fit <- nnetar(window(lynx,end=1920), decay=0.5, maxit=150) plot(forecast(fit,h=14))
## Apply fitted model to later data, including all optional arguments fit2 <- nnetar(window(lynx,start=1921), model=fit)