Merge pull request #310 from tidymodels/a_little_unbalanced

Remove balance

Author: juliasilge

NEWS.md

@@ -16,8 +16,6 @@
 
 * Added better printing methods for initial split objects.
 
-* Added a new `balance` option to `group_vfold_cv()` to balance folds either by the number of groups or the number of observations (@mikemahoney218, #300).
-
 # rsample 0.1.1
 
 * Updated documentation on stratified sampling (#245).

R/groups.R

@@ -22,10 +22,9 @@
 #'
 #' set.seed(123)
 #' group_vfold_cv(Sacramento, group = city, v = 5)
-#' group_vfold_cv(Sacramento, group = city, v = 5, balance = "observations")
 #'
 #' @export
-group_vfold_cv <- function(data, group = NULL, v = NULL, balance = c("groups", "observations"), ...) {
+group_vfold_cv <- function(data, group = NULL, v = NULL, ...) {
   if (!missing(group)) {
     group <- tidyselect::vars_select(names(data), !!enquo(group))
     if (length(group) == 0) {
@@ -42,9 +41,7 @@ group_vfold_cv <- function(data, group = NULL, v = NULL, balance = c("groups", "
     rlang::abort("`group` should be a column in `data`.")
   }
 
-  balance <- rlang::arg_match(balance)
-
-  split_objs <- group_vfold_splits(data = data, group = group, v = v, balance = balance)
+  split_objs <- group_vfold_splits(data = data, group = group, v = v)
 
   ## We remove the holdout indices since it will save space and we can
   ## derive them later when they are needed.
@@ -68,9 +65,7 @@ group_vfold_cv <- function(data, group = NULL, v = NULL, balance = c("groups", "
   )
 }
 
-group_vfold_splits <- function(data, group, v = NULL, balance = c("groups", "observations")) {
-
-  balance <- rlang::arg_match(balance)
+group_vfold_splits <- function(data, group, v = NULL) {
 
   group <- getElement(data, group)
   max_v <- length(unique(group))
@@ -81,7 +76,7 @@ group_vfold_splits <- function(data, group, v = NULL, balance = c("groups", "obs
     check_v(v = v, max_v = max_v, rows = "rows", call = rlang::caller_env())
   }
 
-  indices <- make_groups(data, group, v, balance)
+  indices <- make_groups(data, group, v)
   indices <- lapply(indices, default_complement, n = nrow(data))
   split_objs <-
     purrr::map(indices,

R/make_groups.R

@@ -8,22 +8,13 @@
 #' @param group A variable in `data` (single character or name) used for
 #'  grouping observations with the same value to either the analysis or
 #'  assessment set within a fold.
-#' @param balance If `v` is less than the number of unique groups, how should
-#'  groups be combined into folds? If `"groups"`, the default, then groups are
-#'  combined randomly to balance the number of groups in each fold.
-#'  If `"observations"`, then groups are combined to balance the number of
-#'  observations in each fold.
 #'
 #' @keywords internal
-make_groups <- function(data, group, v, balance) {
+make_groups <- function(data, group, v) {
   data_ind <- data.frame(..index = 1:nrow(data), ..group = group)
   data_ind$..group <- as.character(data_ind$..group)
 
-  res <- switch(
-    balance,
-    "groups" = balance_groups(data_ind, v),
-    "observations" = balance_observations(data_ind, v)
-  )
+  res <- balance_groups(data_ind, v)
   data_ind <- res$data_ind
   keys <- res$keys
 
@@ -47,27 +38,3 @@ balance_groups <- function(data_ind, v) {
     keys = keys
   )
 }
-
-balance_observations <- function(data_ind, v) {
-  while (vec_unique_count(data_ind$..group) > v) {
-    freq_table <- vec_count(data_ind$..group)
-    # Recategorize the largest group to be collapsed
-    # as the smallest group to be kept
-    group_to_keep <- vec_slice(freq_table, v)
-    group_to_collapse <- vec_slice(freq_table, v + 1)
-    collapse_lgl <- vec_in(data_ind$..group, group_to_collapse$key)
-    vec_slice(data_ind$..group, collapse_lgl) <- group_to_keep$key
-  }
-  unique_groups <- unique(data_ind$..group)
-
-  keys <- data.frame(
-    ..group = unique_groups,
-    ..folds = sample(rep(seq_len(v), length.out = length(unique_groups)))
-  )
-
-  list(
-    data_ind = data_ind,
-    keys = keys
-  )
-
-}

man/group_vfold_cv.Rd

@@ -84,37 +84,6 @@ test_that("tibble input", {
   expect_true(all(table(sp_out) == 1))
 })
 
-test_that("other balance methods", {
-  data(ames, package = "modeldata")
-  set.seed(11)
-  rs1 <- group_vfold_cv(ames, "Neighborhood", balance = "observations", v = 2)
-  sizes1 <- dim_rset(rs1)
-
-  expect_true(all(sizes1$analysis == 1465))
-  expect_true(all(sizes1$assessment == 1465))
-  same_data <-
-    purrr::map_lgl(rs1$splits, function(x) {
-      all.equal(x$data, ames)
-    })
-  expect_true(all(same_data))
-
-  good_holdout <- purrr::map_lgl(
-    rs1$splits,
-    function(x) {
-      length(intersect(x$in_ind, x$out_id)) == 0
-    }
-  )
-  expect_true(all(good_holdout))
-
-  expect_true(
-    !any(
-      unique(as.character(assessment(rs1$splits[[1]])$Neighborhood)) %in%
-        unique(as.character(analysis(rs1$splits[[1]])$Neighborhood))
-    )
-  )
-
-})
-
 test_that("printing", {
   expect_snapshot(group_vfold_cv(warpbreaks, "tension"))
 })