A recipe is a description of the steps to be applied to a data set in order to prepare it for data analysis.
recipe(x, ...)# S3 method for default
recipe(x, ...)
# S3 method for data.frame
recipe(x, formula = NULL, ..., vars = NULL, roles = NULL)
# S3 method for formula
recipe(formula, data, ...)
# S3 method for matrix
recipe(x, ...)
An object of class recipe with sub-objects:
A tibble containing information about the original data set columns
A tibble that contains the current set of terms in the
data set. This initially defaults to the same data contained in
var_info.
A list of step or check objects that define the sequence of
preprocessing operations that will be applied to data. The default value is
NULL
A tibble of the data. This is initialized to be the same
as the data given in the data argument but can be different after
the recipe is trained.
A data frame or tibble of the template data set (see below).
Further arguments passed to or from other methods (not currently used).
A model formula. No in-line functions should be used here
(e.g. log(x), x:y, etc.) and minus signs are not allowed. These types of
transformations should be enacted using step functions in this package.
Dots are allowed as are simple multivariate outcome terms (i.e. no need for
cbind; see Examples). A model formula may not be the best choice for
high-dimensional data with many columns, because of problems with memory.
A character string of column names corresponding to variables that will be used in any context (see below)
A character string (the same length of vars) that
describes a single role that the variable will take. This value could be
anything but common roles are "outcome", "predictor",
"case_weight", or "ID"
Variables in recipes can have any type of role, including outcome, predictor, observation ID, case weights, stratification variables, etc.
recipe objects can be created in several ways. If an analysis only
contains outcomes and predictors, the simplest way to create one is to
use a formula (e.g. y ~ x1 + x2) that does not contain inline
functions such as log(x3) (see the first example below).
Alternatively, a recipe object can be created by first specifying
which variables in a data set should be used and then sequentially
defining their roles (see the last example). This alternative is an
excellent choice when the number of variables is very high, as the
formula method is memory-inefficient with many variables.
There are two different types of operations that can be sequentially added to a recipe.
Steps can include operations like scaling a variable, creating dummy variables or interactions, and so on. More computationally complex actions such as dimension reduction or imputation can also be specified.
Checks are operations that conduct specific tests of the data. When the test is satisfied, the data are returned without issue or modification. Otherwise, an error is thrown.
If you have defined a recipe and want to see which steps are included,
use the tidy() method on the recipe object.
Note that the data passed to recipe() need not be the
complete data that will be used to train the steps (by
prep()). The recipe only needs to know the names and types
of data that will be used. For large data sets, head() could
be used to pass a smaller data set to save time and memory.
Once a recipe is defined, it needs to be estimated before being
applied to data. Most recipe steps have specific quantities that must be
calculated or estimated. For example,
step_normalize() needs to compute the training
set’s mean for the selected columns, while
step_dummy() needs to determine the factor levels of
selected columns in order to make the appropriate indicator columns.
The two most common application of recipes are modeling and stand-alone preprocessing. How the recipe is estimated depends on how it is being used.
The best way to use use a recipe for modeling is via the workflows
package. This bundles a model and preprocessor (e.g. a recipe) together
and gives the user a fluent way to train the model/recipe and make
predictions.
library(dplyr)
library(workflows)
library(recipes)
library(parsnip)data(biomass, package = "modeldata")
# split data
biomass_tr <- biomass[biomass$dataset == "Training",]
biomass_te <- biomass[biomass$dataset == "Testing",]
# With only predictors and outcomes, use a formula:
rec <- recipe(HHV ~ carbon + hydrogen + oxygen + nitrogen + sulfur,
data = biomass_tr)
# Now add preprocessing steps to the recipe:
sp_signed <-
rec %>%
step_normalize(all_numeric_predictors()) %>%
step_spatialsign(all_numeric_predictors())
sp_signed
## Recipe
##
## Inputs:
##
## role #variables
## outcome 1
## predictor 5
##
## Operations:
##
## Centering and scaling for all_numeric_predictors()
## Spatial sign on all_numeric_predictors()
We can create a parsnip model, and then build a workflow with the
model and recipe:
linear_mod <- linear_reg()linear_sp_sign_wflow <-
workflow() %>%
add_model(linear_mod) %>%
add_recipe(sp_signed)
linear_sp_sign_wflow
## == Workflow ==========================================================
## Preprocessor: Recipe
## Model: linear_reg()
##
## -- Preprocessor ------------------------------------------------------
## 2 Recipe Steps
##
## * step_normalize()
## * step_spatialsign()
##
## -- Model -------------------------------------------------------------
## Linear Regression Model Specification (regression)
##
## Computational engine: lm
To estimate the preprocessing steps and then fit the linear model, a
single call to [fit()][fit.model_spec()] is used:
linear_sp_sign_fit <- fit(linear_sp_sign_wflow, data = biomass_tr)
When predicting, there is no need to do anything other than call
[predict()][predict.model_fit()]. This preprocesses the new data
in the same manner as the training set, then gives the data to the
linear model prediction code:
predict(linear_sp_sign_fit, new_data = head(biomass_te))
## # A tibble: 6 x 1
## .pred
## <dbl>
## 1 18.1
## 2 17.9
## 3 17.2
## 4 18.8
## 5 19.6
## 6 14.6
When using a recipe to generate data for a visualization or to troubleshoot any problems with the recipe, there are functions that can be used to estimate the recipe and apply it to new data manually.
Once a recipe has been defined, the prep() function can be
used to estimate quantities required for the operations using a data set
(a.k.a. the training data). prep() returns a recipe.
As an example of using PCA (perhaps to produce a plot):
# Define the recipe
pca_rec <-
rec %>%
step_normalize(all_numeric_predictors()) %>%
step_pca(all_numeric_predictors())
Now to estimate the normalization statistics and the PCA loadings:
pca_rec <- prep(pca_rec, training = biomass_tr)
pca_rec
## Recipe
##
## Inputs:
##
## role #variables
## outcome 1
## predictor 5
##
## Training data contained 456 data points and no missing data.
##
## Operations:
##
## Centering and scaling for carbon, hydrogen, oxygen, nitrogen, s... [trained]
## PCA extraction with carbon, hydrogen, oxygen, nitrogen, su... [trained]
Note that the estimated recipe shows the actual column names captured by the selectors.
You can tidy.recipe() a recipe, either when it is
prepped or unprepped, to learn more about its components.
tidy(pca_rec)
## # A tibble: 2 x 6
## number operation type trained skip id
## <int> <chr> <chr> <lgl> <lgl> <chr>
## 1 1 step normalize TRUE FALSE normalize_AeYA4
## 2 2 step pca TRUE FALSE pca_Zn1yz
You can also tidy() recipe steps with a number
or id argument.
To apply the prepped recipe to a data set, the [bake()] function is
used in the same manner that [predict()][predict.model_fit()]
would be for models. This applies the estimated steps to any data set.
bake(pca_rec, head(biomass_te))
## # A tibble: 6 x 6
## HHV PC1 PC2 PC3 PC4 PC5
## <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 18.3 0.730 0.412 0.495 0.333 0.253
## 2 17.6 0.617 -1.41 -0.118 -0.466 0.815
## 3 17.2 0.761 -1.10 0.0550 -0.397 0.747
## 4 18.9 0.0400 -0.950 -0.158 0.405 -0.143
## 5 20.5 0.792 0.732 -0.204 0.465 -0.148
## 6 18.5 0.433 0.127 0.354 -0.0168 -0.0888
In general, the workflow interface to recipes is recommended for most applications.
[biomass$dataset == "Training",]: R:biomass$dataset%20==%20%22Training%22,
[biomass$dataset == "Testing",]: R:biomass$dataset%20==%20%22Testing%22,
[fit.model_spec()]: R:fit.model_spec() [predict.model_fit()]:
R:predict.model_fit() prep(): R:prep() prep():
R:prep() tidy.recipe(): R:tidy.recipe()
tidy.recipe(): R:tidy.recipe() [bake()]: R:bake()
[predict.model_fit()]: R:predict.model_fit()
# formula example with single outcome:
data(biomass, package = "modeldata")
# split data
biomass_tr <- biomass[biomass$dataset == "Training", ]
biomass_te <- biomass[biomass$dataset == "Testing", ]
# With only predictors and outcomes, use a formula
rec <- recipe(
HHV ~ carbon + hydrogen + oxygen + nitrogen + sulfur,
data = biomass_tr
)
# Now add preprocessing steps to the recipe
sp_signed <- rec %>%
step_normalize(all_numeric_predictors()) %>%
step_spatialsign(all_numeric_predictors())
sp_signed
# ---------------------------------------------------------------------------
# formula multivariate example:
# no need for `cbind(carbon, hydrogen)` for left-hand side
multi_y <- recipe(carbon + hydrogen ~ oxygen + nitrogen + sulfur,
data = biomass_tr
)
multi_y <- multi_y %>%
step_center(all_numeric_predictors()) %>%
step_scale(all_numeric_predictors())
# ---------------------------------------------------------------------------
# example using `update_role` instead of formula:
# best choice for high-dimensional data
rec <- recipe(biomass_tr) %>%
update_role(carbon, hydrogen, oxygen, nitrogen, sulfur,
new_role = "predictor"
) %>%
update_role(HHV, new_role = "outcome") %>%
update_role(sample, new_role = "id variable") %>%
update_role(dataset, new_role = "splitting indicator")
rec
Run the code above in your browser using DataLab