Cross-Validated Random Forest MSE for a Sample Design

Trains a Random Forest with k-fold cross-validation on a candidate sample design and returns the mean squared error for a chosen response covariate. Used as the objective function for ss_rf_optimize().

Usage

ss_rf_mse(sample_points, full_data = NULL, n_folds = 5, target_var = NULL)

Arguments

sample_points

Data frame with x, y as the first two columns followed by covariate columns (as returned by ss_clhs_sample()).

full_data

Optional data frame of the full covariate population; if supplied, sample_points is restricted to the shared columns.

n_folds

Integer, number of cross-validation folds. Default 5.

target_var

Character, name of the covariate to predict. If NULL (default), the third covariate column is used, or the first if fewer than three are available.

Value

Numeric, the mean cross-validated MSE, or NA_real_ if the design has too few rows/columns or every fold fails.

See also

ss_rf_optimize()

Examples

if (FALSE) { # \dontrun{
samples <- ss_clhs_sample(r, n_samples = 100, seed = 1)
ss_rf_mse(samples)
} # }