CODEWITHSUNDEEP
rr-caret
chen
chen

Reputation: 69

Using Caret to find the important of features

I got a problem while using Caret.

I used the following code to build model in nnet method, it works.

model_nnet<-train(trainSetSmall[,predictors],trainSetSmall[,outcomeName],method='nnet')
importance <- varImp(model_nnet, scale=FALSE)
plot(importance)

Then, I want to try other models. I tried "gbm", but it does not work.

model_gbm<-train(trainSetSmall[,predictors],trainSetSmall[,outcomeName],method='gbm')
importance2 <- varImp(model_gbm, scale=FALSE)

Error Message:  > importance2 <- varImp(model_gbm, scale=FALSE)
Error in relative.influence(object, n.trees = numTrees) : 
could not find function "relative.influence"

I tried others models also, but except nnet, I canot find another worked one. Sample of My Data (All columns are numeric):

structure(list(x_NMI = c(6347, 6347), EstimateReadBitmaskInd = c(0, 
0), MeterRegActiveReadingDt = c("15-01-2013", "18-01-2013"), 
    MtrRegActNetEngyDailyKwh = c(16.736, 18.093), MtrRgActNetEngyMaxdlyKwh = c(0.831, 
    0.65), RegisterId = c(2, 2), RegisterType = c(2, 2), Building = c(6, 
    6), numberofpeople = c(5, 5), pool = c(2, 2), typeofAC = c(1, 
    1), NoOfAc = c(1, 1)), row.names = 1:2, class = "data.frame")

Upvotes: 1

Views: 508

Answers (1)

josephsturm
josephsturm

Reputation: 323

Looks like relative.influence is a function in the package gbm as can be seen here:

relative.influence documentation

I installed caret and tried to train a model on some trivial data using gbm. This message prompted me to install the gbm package.

enter image description here

Once I installed the gbm package, I was able to train models using gbm in caret.

Upvotes: 1

Related Questions