One standard-error rule selection helper

Aggregates cross-validation scores by parameter setting and returns both the best-performing parameter and the one within one standard error of the best.

Usage

dkge_one_se(scores, param_col = "param", metric_col = "score")

Arguments

scores

Data frame containing per-fold scores.

param_col

Column name identifying the tuning parameter.

metric_col

Column name holding the metric (larger is better).

Value

List with `best`, `pick`, and `summary` table of mean/se by parameter.

Examples

scores <- data.frame(param = rep(1:3, each = 2), score = c(0.2, 0.1, 0.25, 0.2, 0.24, 0.23))
dkge_one_se(scores, param_col = "param", metric_col = "score")$pick
#> [1] 3