Skip to contents

Control Parameters for Racing Survival Workflows

Source: R/racing_workflows.R
control_race_survival.Rd

Creates a racing control specification via control_race with default settings optimized for clinical survival workflows and nested cross-validation.

Usage

control_race_survival(
  save_pred = TRUE,
  parallel_over = c("everything", "resamples", "across"),
  save_workflow = TRUE,
  burn_in = 3,
  num_ties = 10,
  alpha = 0.05,
  randomize = TRUE,
  ...
)

Arguments

save_pred

Logical; whether to save out-of-fold assessment predictions (default TRUE).

parallel_over

How to parallelize execution: "everything" (default), "resamples", or "across".

save_workflow

Logical; whether to retain the fitted workflow object in the output (default TRUE), required by fit_best.

burn_in

Minimum number of resamples evaluated before candidate configurations can be eliminated (default 3).

num_ties

Number of bootstrap samples used to evaluate ties in win-fraction racing (default 10).

alpha

Significance level threshold for elimination in ANOVA racing (default 0.05).

randomize

Logical; whether to randomize the order of resamples (default TRUE).

...

Additional arguments forwarded to control_race.

Value

A control_race object.

See also

tune_race_survival, control_race

Examples

if (FALSE) { # \dontrun{
if (requireNamespace("finetune", quietly = TRUE)) {
  ctrl <- control_race_survival()
}
} # }