Extract Variable Importance from a Maxent Model Object — maxent_variable

This function extracts the percent contribution and permutation importance for each variable from a fitted Maxent model (class "MaxEnt") as returned by dismo::maxent(). It returns a tibble with variables and their respective importance metrics.

Usage

maxent_variable_importance(model = NULL)

Arguments

model: A fitted Maxent model object of class "MaxEnt" (from dismo::maxent().

Value

A tibble with columns:

variable: Variable name (character).
percent_contribution: Percent contribution of the variable (numeric).
permutation_importance: Permutation importance of the variable (numeric).

Author

Ahmed El-Gabbas

Examples

require(ecokit)
ecokit::load_packages(fs, dismo, rJava, raster)

# fit a Maxent model
if (dismo::maxent(silent = TRUE)) {
  predictors <- list.files(
    path = fs::path(
      system.file(package = "dismo"), "ex"),
    pattern = "grd", full.names = TRUE) %>%
    raster::stack()

  occurence <- fs::path(
    system.file(package = "dismo"), "ex", "bradypus.csv") %>%
    read.table(header = TRUE, sep = ",") %>%
    dplyr::select(-1)
  # fit model, biome is a categorical variable
  me <- maxent(predictors, occurence, factors='biome')

  maxent_variable_importance(me)
}
#> # A tibble: 9 × 3
#>   variable percent_contribution permutation_importance
#>   <chr>                   <dbl>                  <dbl>
#> 1 bio1                    2.89                    3.59
#> 2 bio12                   1.69                   12.8 
#> 3 bio16                  10.1                     2.11
#> 4 bio17                   3.57                    4.29
#> 5 bio5                    3.53                    5.36
#> 6 bio6                    3.00                    3.19
#> 7 bio7                   28.2                    54.6 
#> 8 bio8                    0.303                   2.40
#> 9 biome                  46.8                    11.6