feature_selection_RF: Perform feature selection on RFs

In courtiol/rangeRinPA: Protected Area Personnel and Ranger Numbers are Insufficient to Deliver Global Expectations

Perform feature selection on RFs

Description

Perform feature selection on RFs

Usage

feature_selection_RF(
  full_fit,
  data,
  rerank = TRUE,
  metric = "RMSE",
  minimise = TRUE,
  rep = 10,
  Ncpu = 1,
  target = "staff_rangers_log",
  seed = 123,
  ...
)

feature_selection_RF_internal(
  full_fit,
  data,
  rerank = TRUE,
  rep = 10,
  Ncpu = 1,
  target = "staff_rangers_log",
  spatial = TRUE,
  seed = 123,
  ...
)

Arguments

full_fit	a full fitted model
data	the full dataset
rerank	whether or not to recompute variable importance recursively during selection (default = TRUE)
metric	the metric used for computing prediction accuracy (see compute_metrics())
minimise	whether the metric should be minimise (TRUE, default) or maximise (FALSE)
rep	the number of cross validation replicates (default = 10)
Ncpu	the number of CPU cores to be used (default = 1)
target	the name of the response variable
seed	the seed used to control the reproducibility of the cross validation
...	additional parameters to be passed to spaMM::fitme()
spatial	either FALSE (default) or TRUE

Functions

• feature_selection_RF(): wrapper function performing the feature selection on RFs with and without the spatial terms

• feature_selection_RF_internal(): internal function performing the feature selection on RFs

courtiol/rangeRinPA documentation built on Sept. 29, 2022, 9:54 a.m.