Skip to contents

Coerce Historical or Tidymodels Tuning Outputs to Standardized Schema (G5)

Source: R/as_tuning_results.R
as_tuning_results.Rd

Adapts older legacy tuning outputs (`tuning_results_df_*.rds`), tidymodels tuning results, or generic hyperparameter evaluation tables to the standardized schema required by [compare_tuning_runs()], [select_candidate_models()], and [plot.cbe_nested_survival_run()].

Usage

as_tuning_results(x, ...)

Arguments

x

A data frame, tibble, path to an `.rds` file, or a tuning result object.

...

Unused.

Value

A standardized [tibble::tibble()] guaranteed to contain `alpha` and `IBS` columns.

Details

Maps historical and tidymodels column synonyms to canonical names:

  • Integrated Brier Score: `ibs`, `mean_ibs`, `mean` (when `.metric` is `"brier_survival_integrated"`), `brier_score` -> `IBS`.

  • Elastic net mixing: `alphas`, `mixture` -> `alpha`.

  • Penalty parameter: `penalty`, `lambdas` -> `lambda`.

  • Model formula: `formulas`, `formula_str` -> `formula`.

  • Resampling split: `split`, `splits`, `resample`, `id` -> `inner_resamples_splits`.

  • Outer cross-validation fold: `fold`, `outer_id` -> `outer_fold`.

See also

[compare_tuning_runs()], [select_candidate_models()]

Examples

# Historical legacy format with lowercase column names
old_df <- data.frame(alphas = c(0.1, 0.5), ibs = c(0.12, 0.10), penalty = c(0.01, 0.02))
standardized <- as_tuning_results(old_df)
names(standardized)
#> [1] "alphas"    "ibs"       "penalty"   "IBS"       "alpha"     "lambda"   
#> [7] ".metric"   ".estimate"