default_xy_blueprint: Default XY blueprint

View source: R/blueprint-xy-default.R

default_xy_blueprintR Documentation

Default XY blueprint


This pages holds the details for the XY preprocessing blueprint. This is the blueprint used by default from mold() if x and y are provided separately (i.e. the XY interface is used).


  intercept = FALSE,
  allow_novel_levels = FALSE,
  composition = "tibble"

## S3 method for class 'data.frame'
mold(x, y, ..., blueprint = NULL)

## S3 method for class 'matrix'
mold(x, y, ..., blueprint = NULL)



A logical. Should an intercept be included in the processed data? This information is used by the process function in the mold and forge function list.


A logical. Should novel factor levels be allowed at prediction time? This information is used by the clean function in the forge function list, and is passed on to scream().


Either "tibble", "matrix", or "dgCMatrix" for the format of the processed predictors. If "matrix" or "dgCMatrix" are chosen, all of the predictors must be numeric after the preprocessing method has been applied; otherwise an error is thrown.


A data frame or matrix containing the predictors.


A data frame, matrix, or vector containing the outcomes.


Not used.


A preprocessing blueprint. If left as NULL, then a default_xy_blueprint() is used.


As documented in standardize(), if y is a vector, then the returned outcomes tibble has 1 column with a standardized name of ".outcome".

The one special thing about the XY method's forge function is the behavior of outcomes = TRUE when a vector y value was provided to the original call to mold(). In that case, mold() converts y into a tibble, with a default name of .outcome. This is the column that forge() will look for in new_data to preprocess. See the examples section for a demonstration of this.


For default_xy_blueprint(), an XY blueprint.


When mold() is used with the default xy blueprint:

  • It converts x to a tibble.

  • It adds an intercept column to x if intercept = TRUE.

  • It runs standardize() on y.


When forge() is used with the default xy blueprint:

  • It calls shrink() to trim new_data to only the required columns and coerce new_data to a tibble.

  • It calls scream() to perform validation on the structure of the columns of new_data.

  • It adds an intercept column onto new_data if intercept = TRUE.


# ---------------------------------------------------------------------------
# Setup

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

train_x <- train["Sepal.Length"]
train_y <- train["Species"]

test_x <- test["Sepal.Length"]
test_y <- test["Species"]

# ---------------------------------------------------------------------------
# XY Example

# First, call mold() with the training data
processed <- mold(train_x, train_y)

# Then, call forge() with the blueprint and the test data
# to have it preprocess the test data in the same way
forge(test_x, processed$blueprint)

# ---------------------------------------------------------------------------
# Intercept

processed <- mold(train_x, train_y, blueprint = default_xy_blueprint(intercept = TRUE))

forge(test_x, processed$blueprint)

# ---------------------------------------------------------------------------
# XY Method and forge(outcomes = TRUE)

# You can request that the new outcome columns are preprocessed as well, but
# they have to be present in `new_data`!

processed <- mold(train_x, train_y)

# Can't do this!
try(forge(test_x, processed$blueprint, outcomes = TRUE))

# Need to use the full test set, including `y`
forge(test, processed$blueprint, outcomes = TRUE)

# With the XY method, if the Y value used in `mold()` is a vector,
# then a column name of `.outcome` is automatically generated.
# This name is what forge() looks for in `new_data`.

# Y is a vector!
y_vec <- train_y$Species

processed_vec <- mold(train_x, y_vec)

# This throws an informative error that tell you
# to include an `".outcome"` column in `new_data`.
try(forge(iris, processed_vec$blueprint, outcomes = TRUE))

test2 <- test
test2$.outcome <- test2$Species
test2$Species <- NULL

# This works, and returns a tibble in the $outcomes slot
forge(test2, processed_vec$blueprint, outcomes = TRUE)

# ---------------------------------------------------------------------------
# Matrix output for predictors

# You can change the `composition` of the predictor data set
bp <- default_xy_blueprint(composition = "dgCMatrix")
processed <- mold(train_x, train_y, blueprint = bp)

hardhat documentation built on March 31, 2023, 10:21 p.m.