Generate forecasts for TFT models

forecast can only be used if the model object doesn't include known predictors that must exist in the data. It's fine if a recipe passed to tft_dataset_spec() computes known predictors though.

Usage

# S3 method for tft_result
forecast(object, horizon = NULL, past_data = NULL)

Arguments

object

The tft_result object that will be used to create predictions.

horizon

Number of time steps ahead to generate predictions.

past_data

If NULL then the data the model was trained on is used. Predictions are made for the period right after past_data.