Skip to contents

Variable Importance

Source: R/var_imp.R
var_imp.Rd

Methods to calculate variable importance for different classifiers.

Usage

var_imp(mod, data, ...)

# Default S3 method
var_imp(mod, data, ...)

# S3 method for class 'randomForest'
var_imp(mod, data, ...)

# S3 method for class 'cv.glmnet'
var_imp(mod, data, ...)

# S3 method for class 'xgb.Booster'
var_imp(mod, data, ...)

# S3 method for class 'nnet'
var_imp(mod, data, ...)

# S3 method for class 'train'
var_imp(mod, data, ...)

# S3 method for class 'svm'
var_imp(mod, data, ...)

Arguments

mod

model object from classification()

data

data frame with rows as samples, columns as features

...

additional arguments to be passed to or from methods

Details

Currently, variable importance methods are implemented for these classifiers:

  • "rf"

  • "xgboost",

  • "mlr_ridge", "mlr_lasso"

  • "svm"

  • "nnet"

Author

Derek Chiu

Examples

data(hgsc)
class <- attr(hgsc, "class.true")
mod <- classification(hgsc, class, "xgboost")
var_imp(mod)
#> # A tibble: 91 × 2
#>    Variable Importance
#>    <chr>         <dbl>
#>  1 VCAN         0.119 
#>  2 CXCL10       0.109 
#>  3 COL5A1       0.0935
#>  4 HLA-DPA1     0.0766
#>  5 MARCKS       0.0720
#>  6 ITGB2        0.0433
#>  7 LCN2         0.0239
#>  8 STEAP3       0.0231
#>  9 STMN1        0.0220
#> 10 CLU          0.0209
#> # ℹ 81 more rows