plug_formula() specifies the terms of the model through the usage of a
formula.
drop_formula() removes the formula as well as any downstream objects
that might get created after the formula is used for preprocessing, such as
terms. Additionally, if the model has already been fit, then the fit is
removed.
replace_formula() first removes the formula, then replaces the previous
formula with the new one. Any model that has already been fit based on this
formula will need to be refit.
plug_formula(x, formula, ..., blueprint = NULL)
drop_formula(x)
replace_formula(x, formula, ..., blueprint = NULL)A tidyflow
A formula specifying the terms of the model. It is advised to not do preprocessing in the formula, and instead use a recipe if that is required.
Not used.
A hardhat blueprint used for fine tuning the preprocessing.
If NULL, hardhat::default_formula_blueprint() is used.
The tidyflow x, updated with either a new or removed formula preprocessor.
To fit a tidyflow, one of plug_formula() or plug_recipe() must be
specified, but not both.
By default tidyflow leaves workflows to figure out which
type of factor/character transformation to happen (either leave factor as is,
transform to N-1 dummies or use a one-hot encoding approach of N dummy columns).
These transformations depend on the specific model supplied in plug_model.
See add_formula for more details on how transformations
are handled.
However, plug_formula allows to override the type of transformation
using the blueprint. For example, by passing
default_formula_blueprint(intercept = TRUE, indicators = "none") to
the blueprint argument of plug_formula you can
enforce that all factors/characters area left without transforming. You can also use
default_formula_blueprint(intercept = TRUE, indicators = "traditional") and
default_formula_blueprint(intercept = TRUE, indicators = "one_hot") to
transform all factors/characters to N-1 dummies or N dummies respectively.
For example, to transform all factors/characters to one-hot encoding, you can pass
the blueprint to plug_formula:
bp <- default_formula_blueprint(intercept = TRUE, indicators = "one_hot")
iris %>%
tidyflow(seed = 21315)
plug_formula(Sepal.Length ~ Species, blueprint = bp) %>%
plug_model(set_engine(parsnip::linear_reg(), "lm")) %>%
fit()For custom transformations between types (for example, applying one-hot on
factors and not on characters), the user can provide a recipe
with a step_dummy step to plug_recipe. See
this vignette
for more details
# Just for the pipe: %>%
library(tibble)
tflow <-
mtcars %>%
tidyflow(seed = 652341) %>%
plug_formula(mpg ~ .)
tflow
#> ══ Tidyflow ════════════════════════════════════════════════════════════════════
#> Data: 32 rows x 11 columns
#> Split: None
#> Formula: mpg ~ .
#> Resample: None
#> Grid: None
#> Model: None
drop_formula(tflow)
#> ══ Tidyflow ════════════════════════════════════════════════════════════════════
#> Data: 32 rows x 11 columns
#> Split: None
#> Recipe/Formula: None
#> Resample: None
#> Grid: None
#> Model: None
replace_formula(tflow, mpg ~ disp)
#> ══ Tidyflow ════════════════════════════════════════════════════════════════════
#> Data: 32 rows x 11 columns
#> Split: None
#> Formula: mpg ~ disp
#> Resample: None
#> Grid: None
#> Model: None