These are the base classes for creating new preprocessing blueprints. All
blueprints inherit from the one created by new_blueprint(), and the default
method specific blueprints inherit from the other three here.
If you want to create your own processing blueprint for a specific method,
generally you will subclass one of the method specific blueprints here. If
you want to create a completely new preprocessing blueprint for a totally new
preprocessing method (i.e. not the formula, xy, or recipe method) then
you should subclass new_blueprint().
new_formula_blueprint(
mold,
forge,
intercept = FALSE,
allow_novel_levels = FALSE,
ptypes = NULL,
formula = NULL,
indicators = "traditional",
composition = "tibble",
...,
subclass = character()
)new_recipe_blueprint(
mold,
forge,
intercept = FALSE,
allow_novel_levels = FALSE,
fresh = TRUE,
bake_dependent_roles = character(),
composition = "tibble",
ptypes = NULL,
recipe = NULL,
...,
subclass = character()
)
new_xy_blueprint(
mold,
forge,
intercept = FALSE,
allow_novel_levels = FALSE,
composition = "tibble",
ptypes = NULL,
...,
subclass = character()
)
new_blueprint(
mold,
forge,
intercept = FALSE,
allow_novel_levels = FALSE,
composition = "tibble",
ptypes = NULL,
...,
subclass = character()
)
A named list with two elements, clean and process, see
the new_blueprint() section, Mold Functions, for details.
A named list with two elements, clean and process, see
the new_blueprint() section, Forge Functions, for details.
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 NULL, or a named list with 2 elements, predictors
and outcomes, both of which are 0-row tibbles. ptypes is generated
automatically at mold() time and is used to validate new_data at
prediction time. At mold() time, the information found in
blueprint$mold$process()$ptype is used to set ptypes for the blueprint.
Either NULL, or a formula that specifies how the
predictors and outcomes should be preprocessed. This argument is set
automatically at mold() time.
A single character string. Control how factors are expanded into dummy variable indicator columns. One of:
"traditional" - The default. Create dummy variables using the
traditional model.matrix() infrastructure. Generally this creates
K - 1 indicator columns for each factor, where K is the number of
levels in that factor.
"none" - Leave factor variables alone. No expansion is done.
"one_hot" - Create dummy variables using a one-hot encoding approach
that expands unordered factors into all K indicator columns, rather than
K - 1.
Either "tibble", "matrix", or "dgCMatrix" for the format of the processed predictors. If "matrix" or "dgCMatrix" are chosen, all of the predictors must be numeric after the preprocessing method has been applied; otherwise an error is thrown.
Name-value pairs for additional elements of blueprints that subclass this blueprint.
A character vector. The subclasses of this blueprint.
Should already trained operations be re-trained when prep() is
called?
A character vector of recipes column "roles"
specifying roles that are required to recipes::bake() new data. Can't be
"predictor" or "outcome", as predictors are always required and
outcomes are handled by the outcomes argument of forge().
Typically, non-standard roles (such as "id" or "case_weights") are not
required to bake() new data. Unless specified by bake_dependent_roles,
these non-standard role columns are excluded from checks done in forge()
to validate the column structure of new_data, will not be passed to
bake() even if they existed in new_data, and will not be returned in
the forge()$extras$roles slot. See the documentation of
recipes::add_role() for more information about roles.
Either NULL, or an unprepped recipe. This argument is set
automatically at mold() time.
A preprocessing blueprint, which is a list containing the inputs used as arguments to the function, along with a class specific to the type of blueprint being created.
blueprint$mold should be a named list with two elements, both of which
are functions:
clean: A function that performs initial cleaning of the user's input
data to be used in the model.
Arguments:
If this is an xy blueprint, blueprint, x and y.
Otherwise, blueprint and data.
Output: A named list of three elements:
blueprint: The blueprint, returned and potentially updated.
If using an xy blueprint:
x: The cleaned predictor data.
y: The cleaned outcome data.
If not using an xy blueprint:
data: The cleaned data.
process: A function that performs the actual preprocessing of the data.
Arguments:
If this is an xy blueprint, blueprint, x and y.
Otherwise, blueprint and data.
Output: A named list of 5 elements:
blueprint: The blueprint, returned and potentially updated.
predictors: A tibble of predictors.
outcomes: A tibble of outcomes.
ptypes: A named list with 2 elements, predictors and outcomes,
where both elements are 0-row tibbles.
extras: Varies based on the blueprint. If the blueprint has no
extra information, NULL. Otherwise a named list of the
extra elements returned by the blueprint.
Both blueprint$mold$clean() and blueprint$mold$process() will be called,
in order, from mold().
blueprint$forge should be a named list with two elements, both of which
are functions:
clean: A function that performs initial cleaning of new_data:
Arguments:
blueprint, new_data, and outcomes.
Output: A named list of the following elements:
blueprint: The blueprint, returned and potentially updated.
predictors: A tibble containing the cleaned predictors.
outcomes: A tibble containing the cleaned outcomes.
extras: A named list of any extras obtained while cleaning. These
are passed on to the process() function for further use.
process: A function that performs the actual preprocessing of the data
using the known information in the blueprint.
Arguments:
blueprint, new_data, outcomes, extras.
Output: A named list of the following elements:
blueprint: The blueprint, returned and potentially updated.
predictors: A tibble of the predictors.
outcomes: A tibble of the outcomes, or NULL.
extras: Varies based on the blueprint. If the blueprint has no
extra information, NULL. Otherwise a named list of the
extra elements returned by the blueprint.
Both blueprint$forge$clean() and blueprint$forge$process() will be called,
in order, from forge().