Customized Version of {workflowsets}

Wrapper around tidyr::crossing() that creates all combinations of recipes and models.

Usage

rpwf_workflow_set(preprocs, models, costs)

Arguments

preprocs

list or vector of recipes::recipe().

models

list or vector of model spec. Generated by adding set_py_engine() to a model, e.g. parsnip::boost_tree() and parsnip::set_engine().

costs

list or vector of sklearn cost optimization metrics such as "neg_log_loss" and "roc_auc". Check the docs for available values.

Value

tibble that contains a combination of list of recipes, models, and costs.

Examples

d <- mtcars
d$id <- seq_len(nrow(d))
m1 <- parsnip::boost_tree() |>
  parsnip::set_engine("xgboost") |>
  parsnip::set_mode("classification") |>
  set_py_engine(py_module = "xgboost", py_base_learner = "XGBClassifier")
r1 <- d |>
  recipes::recipe(vs ~ .) |>
  # "pd.index" is the special column that used for indexing in pandas
  recipes::update_role(id, new_role = "pd.index")
wf <- rpwf_workflow_set(list(r1), list(m1), "neg_log_loss")
wf
#> # A tibble: 1 × 3
#>   preprocs models    costs       
#> * <list>   <list>    <chr>       
#> 1 <recipe> <spec[+]> neg_log_loss