using getImpXgboost as getImp method seems raise the following traceback:
'Error in ans[as.numeric(imp$Feature) + 1] <- imp$Gain :
NAs are not allowed in subscripted assignments
In addition: Warning message:
In getImp(cbind(x[, decReg != "Rejected"], xSha), y, ...) :
NAs introduced by coercion'
of course, I converted all integer features to numeric,(i tried also to convert dataframe to matrix, but isn't here the problem) but the issue remains the same...what could be the problem?
ps. with default 'getImpRfZ'method all works fine;
Thanks for reporting, this is indeed a bug caused by a change in the xgboost package. Fix is on branch issue2 ( dc857ea488), you can install with
now works great!