borg_vfold_cv: BORG-Guarded vfold_cv

In BORG: Bounded Outcome Risk Guard for Model Evaluation

BORG-Guarded vfold_cv

Description

A guarded version of rsample::vfold_cv() that checks for data dependencies before creating folds. If spatial, temporal, or clustered dependencies are detected, random CV is blocked.

Usage

borg_vfold_cv(
  data,
  v = 10,
  repeats = 1,
  strata = NULL,
  coords = NULL,
  time = NULL,
  groups = NULL,
  target = NULL,
  allow_override = FALSE,
  auto_block = FALSE,
  ...
)

Arguments

data	A data frame.
v	Integer. Number of folds. Default: 10.
repeats	Integer. Number of repeats. Default: 1.
strata	Character. Column name for stratification.
coords	Character vector of length 2. Coordinate columns for spatial check.
time	Character. Time column for temporal check.
groups	Character. Group column for clustered check.
target	Character. Target variable for dependency detection.
allow_override	Logical. If TRUE, allow random CV with explicit confirmation. Default: FALSE.
auto_block	Logical. If TRUE, automatically switch to blocked CV when dependencies detected. If FALSE, throw error. Default: FALSE.
...	Additional arguments passed to rsample::vfold_cv().

Value

If no dependencies detected or allow_override = TRUE, returns an rset object from rsample. If dependencies detected and auto_block = TRUE, returns BORG-generated blocked CV folds.

See Also

borg_cv for direct blocked CV generation.

Examples

if (requireNamespace("rsample", quietly = TRUE)) {
  # Safe: no dependencies
  data <- data.frame(x = rnorm(100), y = rnorm(100))
  folds <- borg_vfold_cv(data, v = 5)

  # Use auto_block to automatically switch to spatial CV:
  spatial_data <- data.frame(
    lon = runif(100, -10, 10),
    lat = runif(100, -10, 10),
    response = rnorm(100)
  )
  folds <- borg_vfold_cv(spatial_data, coords = c("lon", "lat"),
                         target = "response", auto_block = TRUE)
}

BORG documentation built on March 20, 2026, 5:09 p.m.