Skip to contents

Run Random Forest Imputation with Optional Parallelization

run_rf_imputation.Rd

This function performs missing value imputation using the missForest package on a subset of columns in a data frame. Column selection can be explicit or based on a regular expression, using resolve_target_cols internally.

Usage

run_rf_imputation(df, target_cols, control_RF = list())

Arguments

df

A data frame containing the data to be imputed. Only non-QC rows should be passed here.

target_cols

A character vector of column names or a regex string to identify columns to impute.

control_RF

A named list of additional or overriding arguments passed to missForest::missForest(). Also supports an internal control argument n_cores (numeric), used to determine parallelism when parallelize is not "no".

Accepted entries include (see missForest::missForest for details):

mtry

Number of variables randomly sampled at each split.

ntree

Number of trees to grow.

maxiter

Maximum number of imputation iterations.

parallelize

Character string: "no", "variables", or "forests". Enables parallelization.

variablewise

Logical. If TRUE, variable-wise error is returned.

verbose

Logical. If TRUE, progress messages are printed.

replace

Logical. Whether sampling of observations is done with replacement.

classwt, cutoff, strata, sampsize, etc.

Other optional arguments passed to randomForest::randomForest.

n_cores

(Internal) Number of CPU cores to use if parallelize is not "no". This is removed before calling missForest().

Value

A list with two components:

imputed

A data frame of the same shape as the input subset, with imputed values.

oob

A named numeric vector of out-of-bag error estimates for each feature (if available).

Details

The function supports optional parallel execution via missForest's parallelize argument. If n_cores is less than or equal to 1, parallelization is disabled automatically and the imputation runs sequentially.

See also

missForest, resolve_target_cols, register_cluster, unregister_cluster

Examples

if (FALSE) { # \dontrun{
df <- data.frame(a = c(1, NA, 3), b = c(4, 5, NA))

# Default imputation on selected columns
run_rf_imputation(df, target_cols = c("a", "b"))

# Use regex to select columns and customize RF parameters
run_rf_imputation(
    df,
    target_cols = "^a|b$",
    control_RF = list(ntree = 200, mtry = 1, n_cores = 4)
)
} # }