Remove target from test set and back up values in reference table

This function removes the target values from the test set (replacing with NA) and backs up the values in a new variable in the global environment. This is particularly useful to avoid target leakage (i.e. accidentally using the target value during testing).

Usage

remove_target_from_test_and_add_ref_to_env(tt, target, unique_id, ref_name)

Arguments

tt

train-test list (see ttsplit())

target

(character) name of target variable

unique_id

(character) name of unique ID variable

ref_name

(character) name of new variable that stores mapping between ID and target

Value

tt with target replaced with NA in $test, and ref_name added to global environment

Note

This function implements the method described in Preventing Target Leakage (D22 QuantCafé, 2021).

See also

add_target_back_to_test_set_from_ref_table()

Examples

tt <- df |>
   dplyr::mutate(id=1:nrow(df)) |>
   ttsplit() |>
   remove_target_from_test_and_add_ref_to_env("y1", "id", "target_values")
#> [1] "Split df into list with: train, test (proportion train = 0.7)"
#> [1] "Target column y1 replaced with NA in test set"
#> [1] "Added reference table 'target_values' to global environment"