Skip to contents

End-to-End cLHS + Random Forest Sampling Optimization

Source: R/rf_optimization.R
ss_rf_size.Rd

Runs the full two-stage workflow: a cLHS baseline design (ss_clhs_sample()) refined by Random Forest simulated annealing (ss_rf_optimize()), with side-by-side comparison plots and an optional export of CSV/PNG outputs.

Usage

ss_rf_size(
  covariates,
  n_samples = 100,
  n_iterations = 500,
  output_dir = NULL,
  seed = 123,
  target_var = NULL,
  covariate_name = NULL
)

Arguments

covariates

A character path to a .tif file or directory (see ss_load_rasters()), or an already-loaded SpatRaster stack.

n_samples

Integer, number of samples in the design. Default 100.

n_iterations

Integer, simulated annealing iterations. Default 500.

output_dir

Character, directory to write CSV/PNG outputs to. If NULL (default), nothing is written to disk.

seed

Optional integer seed, passed to ss_rf_optimize(). Default 123.

target_var

Character, covariate to use as the RF response. Passed to ss_rf_mse().

covariate_name

Character, background layer for the comparison plots. Passed to the internal plotting helper.

Value

A list with:

clhs_samples

Data frame, the cLHS baseline design.

rf_optimized_samples

Data frame, the RF-optimized design.

comparison_table

Data frame comparing MSE and improvement.

plots

List with clhs_plot and rf_plot.

improvement

Numeric, percent MSE reduction.

file_paths

Only when output_dir is supplied: named list of files written.

See also

ss_clhs_sample(), ss_rf_optimize(), ss_rf_mse()

Examples

if (FALSE) { # \dontrun{
res <- ss_rf_size("data/predictors.tif", n_samples = 100, n_iterations = 500)
res$improvement
} # }