tsfgrnn 1.0.5

• Now, if the lags are automatically chosen, it is not allowed to use only one autoregressive lag with the additive ormultiplicative transformation (it does not make sense)

tsfgrnn 1.0.4

• autoplot.grnnForecast has been modified to comply with CRAN

tsfgrnn 1.0.3

• citation information has been added

tsfgrnn 1.0.2

• References to the paper describing this package have been added in the documentation

tsfgrnn 1.0.1

• The default transformation is now additive instead of multiplicative.

tsfgrnn 1.0.0

• Automatic computation of smoothing parameter can be done with rolling origin or fixed origin (less accurate, but faster).

• Lags can be selected using forward selection or backward elimination (feature selection techniques)

• Time series cannot be scaled

• The model is built using faster Rcpp code

• An optional transformation to the training samples has been added. It improves forecast accuracy for time series with a trend

tsfgrnn 0.2.0

• Uses Rcpp for speeding up computation of forecasts.

tsfgrnn 0.1.0

• This is the first release of this package.