K-fold BRT fit — eval_kfold

k-fold cross validation code from Camrin Brawn (WHOI): https://zenodo.org/records/7971532. Function originally by B. Abrahms from https://github.com/briana-abrahms/DynamicEnsembleSDM/blob/master/model_evaluation.R. Modified to return information necessary to calculate evaluation metrics. Follows many of the same inputs as gbm.step

Usage

eval_kfold_brt(
  data_input,
  gbm_x,
  gbm_y,
  learning_rate = 0.05,
  k_folds = 5,
  tree_complexity = 3,
  bag_fraction = 0.6,
  is_fixed = TRUE,
  max_trees = 2000
)

Arguments

data_input: input data frame
gbm_x: names of predictor variables in data_input
gbm_y: name of response variable in data_input
learning_rate: sets the weight applied to individual trees input to dismo::gbm.step
k_folds: number of folds to perform
tree_complexity: is tree complexity input to dismo::gbm.step - sets complexity of individual trees
bag_fraction: is bag fraction input to dismo::gbm.step - sets the proportion of observations used in selecting variables
is_fixed: TRUE/FALSE to determine if gbm.step or gbm.fixed should be used.
max_trees: maximum number of trees to fit before stopping

Value

a list containing 1) the output from eval_brt and 2) a data.frame of the observed and predicted values used to calculate RMSE/AUC by sdm_cv

Examples

if (FALSE) { # \dontrun{
eval_kfold_brt(data_input_Fit, gbm_x=c("curl","ild", "ssh", "sst","sst_sd"), "presabs")
} # }