Cross-validation for tidylearn models

Usage

tl_cv(data, formula, method, folds = 5, metrics = NULL, transform = NULL, ...)

Arguments

data: Data frame
formula: Model formula
method: Modeling method
folds: Number of cross-validation folds
metrics: Character vector of metrics to compute on each fold, passed to tl_evaluate. If NULL (the default), tl_evaluate's per-task defaults are used.
transform: Optional function for feature engineering that has to be refitted per fold. It is called with the training rows of each fold and must return a list with an apply function (applied to both the training and assessment rows) and, optionally, a formula to fit under. Use this for anything that learns parameters from the data – PCA rotations, cluster centroids, target encodings – since fitting those before the split inflates every fold's score.
...: Additional arguments passed to tl_model

Value

A list with two elements:

$folds: A list of per-fold evaluation tibbles, each with metric and value columns.
$summary: A tibble with columns metric, mean, and sd summarizing performance across folds.

Examples

# \donttest{
cv <- tl_cv(mtcars, mpg ~ wt + hp, method = "linear", folds = 3)
cv$summary
#> # A tibble: 3 × 3
#>   metric  mean    sd
#>   <chr>  <dbl> <dbl>
#> 1 mae    2.09  0.779
#> 2 rmse   2.63  0.927
#> 3 rsq    0.737 0.214
# }