step_classdist: Distances to class centroids
In tidymodels/recipes: Preprocessing and Feature Engineering Steps for Modeling

step_classdist

R Documentation

Distances to class centroids

Description

step_classdist() creates a specification of a recipe step that will convert numeric data into Mahalanobis distance measurements to the data centroid. This is done for each value of a categorical class variable.

Usage

step_classdist(
  recipe,
  ...,
  class,
  role = "predictor",
  trained = FALSE,
  mean_func = mean,
  cov_func = cov,
  pool = FALSE,
  log = TRUE,
  objects = NULL,
  prefix = "classdist_",
  keep_original_cols = TRUE,
  skip = FALSE,
  id = rand_id("classdist")
)

Arguments

`recipe`	A recipe object. The step will be added to the sequence of operations for this recipe.
`...`	One or more selector functions to choose variables for this step. See `selections()` for more details.
`class`	A single character string that specifies a single categorical variable to be used as the class.
`role`	For model terms created by this step, what analysis role should they be assigned? By default, the new columns created by this step from the original variables will be used as predictors in a model.
`trained`	A logical to indicate if the quantities for preprocessing have been estimated.
`mean_func`	A function to compute the center of the distribution.
`cov_func`	A function that computes the covariance matrix
`pool`	A logical: should the covariance matrix be computed by pooling the data for all of the classes?
`log`	A logical: should the distances be transformed by the natural log function?
`objects`	Statistics are stored here once this step has been trained by `prep()`.
`prefix`	A character string for the prefix of the resulting new variables. See notes below.
`keep_original_cols`	A logical to keep the original variables in the output. Defaults to `TRUE`.
`skip`	A logical. Should the step be skipped when the recipe is baked by `bake()`? While all operations are baked when `prep()` is run, some operations may not be able to be conducted on new data (e.g. processing the outcome variable(s)). Care should be taken when using `skip = TRUE` as it may affect the computations for subsequent operations.
`id`	A character string that is unique to this step to identify it.

Details

step_classdist will create a new column for every unique value of the class variable. The resulting variables will not replace the original values and, by default, have the prefix classdist_. The naming format can be changed using the prefix argument.

Class-specific centroids are the multivariate averages of each predictor using the data from each class in the training set. When pre-processing a new data point, this step computes the distance from the new point to each of the class centroids. These distance features can be very effective at capturing linear class boundaries. For this reason, they can be useful to add to an existing predictor set used within a nonlinear model. If the true boundary is actually linear, the model will have an easier time learning the training data patterns.

Note that, by default, the default covariance function requires that each class should have at least as many rows as variables listed in the terms argument. If pool = TRUE, there must be at least as many data points are variables overall.

Value

An updated version of recipe with the new step added to the sequence of any existing operations.

Tidying

When you tidy() this step, a tibble is returned with columns terms, value, class , and id:

terms: character, the selectors or variables selected
value: numeric, location of centroid
class: character, name of the class
id: character, id of this step

Case weights

This step performs an supervised operation that can utilize case weights. As a result, case weights are used with frequency weights as well as importance weights. For more information,, see the documentation in case_weights and the examples on tidymodels.org.

Examples


data(penguins, package = "modeldata")
penguins <- penguins[vctrs::vec_detect_complete(penguins), ]
penguins$island <- NULL
penguins$sex <- NULL

# in case of missing data...
mean2 <- function(x) mean(x, na.rm = TRUE)

# define naming convention
rec <- recipe(species ~ ., data = penguins) %>%
  step_classdist(all_numeric_predictors(),
    class = "species",
    pool = FALSE, mean_func = mean2, prefix = "centroid_"
  )

# default naming
rec <- recipe(species ~ ., data = penguins) %>%
  step_classdist(all_numeric_predictors(),
    class = "species",
    pool = FALSE, mean_func = mean2
  )

rec_dists <- prep(rec, training = penguins)

dists_to_species <- bake(rec_dists, new_data = penguins)
## on log scale:
dist_cols <- grep("classdist", names(dists_to_species), value = TRUE)
dists_to_species[, c("species", dist_cols)]

tidy(rec, number = 1)
tidy(rec_dists, number = 1)

tidymodels/recipes documentation built on Nov. 29, 2024, 3:05 p.m.