tests/testthat/test_mlr_resampling_holdout.R

test_that("holdout has no duplicated ids", {
  r = rsmp("holdout")
  expect_identical(r$duplicated_ids, FALSE)
})

test_that("stratification", {
  data = data.table(y = factor(rep(letters[1:2], times = c(90, 10))), x1 = runif(100), x2 = rep(LETTERS[1:2], times = c(50, 50)))
  b = as_data_backend(data)
  task = TaskClassif$new("stratify_data", b, target = "y")
  task$col_roles$stratum = task$target_names

  r = rsmp("holdout", ratio = .5)
  r$instantiate(task)

  i = 1L
  expect_equal(task$data(r$train_set(i))[y == "a", .N], 45)
  expect_equal(task$data(r$train_set(i))[y == "b", .N], 5)
  expect_equal(task$data(r$test_set(i))[y == "a", .N], 45)
  expect_equal(task$data(r$test_set(i))[y == "b", .N], 5)
})

test_that("grouping", {
  r = rsmp("holdout")
  expect_grouping_works(r)
})

test_that("prediction does not drop dimension (#551)", {
  task = tsk("iris")
  learner = lrn("classif.rpart")
  resampling = rsmp("holdout")
  resampling$instantiate(task)

  design = data.table(
    learner = list(learner),
    task = list(task),
    resampling = list(resampling)
  )

  bmr = benchmark(design)
  expect_number(bmr$aggregate(msr("classif.ce"))[["classif.ce"]])
  expect_equal(map(get_private(bmr)$.data$data$fact$prediction, names), list("test"))
})

Try the mlr3 package in your browser

Any scripts or data that you put into this service are public.

mlr3 documentation built on Nov. 17, 2023, 5:07 p.m.