The forecastHybrid
R package gives a platform to ensemble heterogeneous time series models. The main function that enables this task is the hybridModel
function. The core function provides the option referred to as models
. It takes as input a string of up to six characters, and the characters are representatives of the models: a
for the auto.arima
model, e
for ets
, f
for thetam
, n
denoting nnetar
, s
for stlm
, and finally, t
represents tbats
. Consequently, if we give a character string of ae
to models, it will combine results from the ARIMA and ets models. This is illustrated on the co2
dataset for different combinations of the time series models:
>accuracy(forecast(co2_arima,h=25),x=co2[444:468]) ME RMSE MAE MPE MAPE MASE ACF1 Training set 0.0185 0.283 0.225 0.00541 0.0672 0.211 0.0119 Test set -0.0332 0.349 0.270 -0.00912 0.0742 0.252 NA >AP_Ensemble_02 <- hybridModel(co2_sub,models="ae") Fitting the auto.arima...