Create a BORG Validation Workflow

Wraps the full model evaluation pipeline into a single trackable object: diagnose data dependencies, generate appropriate CV folds, fit a model per fold, and validate each split for leakage.

Usage

borg_workflow(
  data,
  formula,
  coords = NULL,
  time = NULL,
  groups = NULL,
  v = 5,
  fit_fun = stats::lm,
  metric = c("rmse", "mae", "rsq"),
  buffer = NULL,
  parallel = FALSE,
  verbose = FALSE,
  ...
)

Arguments

data

Data frame with predictor and target columns.

formula

Model formula (e.g. y ~ x1 + x2).

coords

Character vector of coordinate column names.

time

Character. Time column name.

groups

Character. Group column name.

v

Integer. Number of folds. Default: 5.

fit_fun

Function to fit a model. Default: lm. Must accept formula and data and return an object with predict().

metric

Character. Performance metric. Default: "rmse".

buffer

Numeric. Optional spatial buffer.

parallel

Logical. If TRUE and future.apply is installed, fold fitting and evaluation run in parallel. Default: FALSE.

verbose

Logical. Default: FALSE.

...

Additional arguments passed to borg_cv().

Value

A borg_workflow object (list) containing:

diagnosis

BorgDiagnosis object

cv

borg_cv object with folds

models

List of fitted models (one per fold)

predictions

List of prediction vectors (one per fold)

performance

borg_fold_perf data frame

risks

List of BorgRisk objects (one per fold)

data

Original data

formula

Model formula

params

Workflow parameters

Examples

set.seed(42)
d <- data.frame(
  x = runif(100, 0, 100), y = runif(100, 0, 100),
  z = rnorm(100)
)
wf <- borg_workflow(d, z ~ x + y, coords = c("x", "y"))
wf