extract predictor names from varImp

Question

I am using the caret package to run a gbm model. Once the model is run, I use the varImp function to extract the list of important predictors (displays top 20). However I would like to capture the names of the predictors in a character list. How do I do that? The object returned from the varImp does not seem to have the attribute that lists the predictor name - only the variable importance. Here is a sample :

gbmModel= train(target ~. , data = trainData, ....other params )
varimp = varImp(gbmModel, scale=TRUE)
str(varimp)
List of 3
 $ importance:'data.frame': 77 obs. of  1 variable:
..$ Overall: num [1:77] 6.63 0 5.35 2.01 0 ...
$ model     : chr "gbm"
$ calledFrom: chr "varImp"
- attr(*, "class")= chr "varImp.train"
## Display the important variables

varimp

gbm variable importance

 only 20 most important variables shown (out of 77)

           Overall
Var126     100.000
Var189      99.647
Var113      41.994
... And so on

I would like a list like ("Var126", "Var189", "Var113" ...)

Thanks in advance,

What about `rownames(varimp$importance)`? Otherwise, please create a [reproducible example](http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) with test data so we can run the same code as you to test possible solutions. — MrFlick, Jun 25 '14 at 15:47
where is `varImp()` coming from? (Not from **gbm**, but perhaps you are using [unstated] the **caret** package?) — Gavin Simpson, Jun 25 '14 at 15:56

score 6 · Answer 1 · answered Jun 25 '14 at 18:28

6

From the structure of the output you provided, it appears that

rownames(varimp$importance)

will return the values you are after.

answered Jun 25 '14 at 18:28

MrFlick

195,160
17
277
295

just `varimp` returns only top 20 features, how to get names of only top 20 predictors ? – parth Jan 30 '18 at 08:54
Anyone knows how to get the names of the predictors instead of the corresponding column number? – user11428 Jan 31 '19 at 15:07
@parth `rownames(varImp(gbmModel)$importance)[1:20]` – C. John May 08 '20 at 17:33

score 1 · Answer 2 · edited Jan 27 '20 at 19:31

1

varImp(gbmModel, scale = TRUE)$importance

Return all variables

edited Jan 27 '20 at 19:31

Dharman

30,962
25
85
135

answered Jan 27 '20 at 19:18

forever

139
1
2
8

extract predictor names from varImp

2 Answers2