Learn R Programming

hardhat (version 1.3.0)

forge: Forge prediction-ready data

Description

forge() applies the transformations requested by the specific blueprint on a set of new_data. This new_data contains new predictors (and potentially outcomes) that will be used to generate predictions.

All blueprints have consistent return values with the others, but each is unique enough to have its own help page. Click through below to learn how to use each one in conjunction with forge().

  • XY Method - default_xy_blueprint()

  • Formula Method - default_formula_blueprint()

  • Recipes Method - default_recipe_blueprint()

Usage

forge(new_data, blueprint, ..., outcomes = FALSE)

Value

A named list with 3 elements:

  • predictors: A tibble containing the preprocessed new_data predictors.

  • outcomes: If outcomes = TRUE, a tibble containing the preprocessed outcomes found in new_data. Otherwise, NULL.

  • extras: Either NULL if the blueprint returns no extra information, or a named list containing the extra information.

Arguments

new_data

A data frame or matrix of predictors to process. If outcomes = TRUE, this should also contain the outcomes to process.

blueprint

A preprocessing blueprint.

...

Not used.

outcomes

A logical. Should the outcomes be processed and returned as well?

Details

If the outcomes are present in new_data, they can optionally be processed and returned in the outcomes slot of the returned list by setting outcomes = TRUE. This is very useful when doing cross validation where you need to preprocess the outcomes of a test set before computing performance.

Examples

Run this code
# See the blueprint specific documentation linked above
# for various ways to call forge with different
# blueprints.

train <- iris[1:100, ]
test <- iris[101:150, ]

# Formula
processed <- mold(
  log(Sepal.Width) ~ Species,
  train,
  blueprint = default_formula_blueprint(indicators = "none")
)

forge(test, processed$blueprint, outcomes = TRUE)

Run the code above in your browser using DataLab