step_upsample is now available as
function creates a specification of a recipe step that
will replicate rows of a data set to make the occurrence of
levels in a specific factor level equal.
step_upsample( recipe, ..., over_ratio = 1, ratio = NA, role = NA, trained = FALSE, column = NULL, target = NA, skip = TRUE, seed = sample.int(10^5, 1), id = rand_id("upsample") ) # S3 method for step_upsample tidy(x, ...)
A recipe object. The step will be added to the sequence of operations for this recipe.
One or more selector functions to choose which
variable is used to sample the data. See
A numeric value for the ratio of the majority-to-minority frequencies. The default value (1) means that all other levels are sampled up to have the same frequency as the most occurring level. A value of 0.5 would mean that the minority levels will have (at most) (approximately) half as many rows than the majority level.
Deprecated argument; same as
Not used by this step since no new variables are created.
A logical to indicate if the quantities for preprocessing have been estimated.
A character string of the variable name that will
be populated (eventually) by the
An integer that will be used to subsample. This
should not be set by the user and will be populated by
A logical. Should the step be skipped when the
recipe is baked by
An integer that will be used as the seed when upsampling.
A character string that is unique to this step to identify it.
An updated version of
recipe with the new step
added to the sequence of existing steps (if any). For the
tidy method, a tibble with columns
terms which is
the variable used to sample.
Up-sampling is intended to be performed on the training set alone. For
this reason, the default is
skip = TRUE. It is advisable to use
prep(recipe, retain = TRUE) when preparing the recipe; in this way
bake(object, new_data = NULL) can be used to obtain the up-sampled version
of the data.
If there are missing values in the factor variable that is used to define the sampling, missing data are selected at random in the same way that the other factor levels are sampled. Missing values are not used to determine the amount of data in the majority level (see example below).
For any data with factor levels occurring with the same frequency as the majority level, all data will be retained.
All columns in the data are sampled and returned by
When used in modeling, users should strongly consider using the
skip = TRUE so that the extra sampling is not
conducted outside of the training set.
library(modeldata) data(okc) orig <- table(okc$diet, useNA = "always") sort(orig, decreasing = TRUE)#> #> <NA> mostly anything anything strictly anything #> 24360 16562 6174 5107 #> mostly vegetarian mostly other strictly vegetarian vegetarian #> 3438 1004 874 665 #> strictly other mostly vegan other strictly vegan #> 450 335 331 227 #> vegan mostly kosher mostly halal strictly halal #> 136 86 48 18 #> strictly kosher halal kosher #> 18 11 11up_rec <- recipe( ~ ., data = okc) %>% # Bring the minority levels up to about 200 each # 200/16562 is approx 0.0121 step_upsample(diet, over_ratio = 0.0121) %>% prep(training = okc)#> Warning: `step_upsample()` is deprecated as of recipes 0.1.13. #> Please use `themis::step_upsample()` instead. #> This warning is displayed once every 8 hours. #> Call `lifecycle::last_warnings()` to see where this warning was generated.training <- table(bake(up_rec, new_data = NULL)$diet, useNA = "always") # Since `skip` defaults to TRUE, baking the step has no effect baked_okc <- bake(up_rec, new_data = okc) baked <- table(baked_okc$diet, useNA = "always") # Note that if the original data contained more rows than the # target n (= ratio * majority_n), the data are left alone: data.frame( level = names(orig), orig_freq = as.vector(orig), train_freq = as.vector(training), baked_freq = as.vector(baked) )#> level orig_freq train_freq baked_freq #> 1 anything 6174 6174 6174 #> 2 halal 11 200 11 #> 3 kosher 11 200 11 #> 4 mostly anything 16562 16562 16562 #> 5 mostly halal 48 200 48 #> 6 mostly kosher 86 200 86 #> 7 mostly other 1004 1004 1004 #> 8 mostly vegan 335 335 335 #> 9 mostly vegetarian 3438 3438 3438 #> 10 other 331 331 331 #> 11 strictly anything 5107 5107 5107 #> 12 strictly halal 18 200 18 #> 13 strictly kosher 18 200 18 #> 14 strictly other 450 450 450 #> 15 strictly vegan 227 227 227 #> 16 strictly vegetarian 874 874 874 #> 17 vegan 136 200 136 #> 18 vegetarian 665 665 665 #> 19 <NA> 24360 24360 24360