factor attributes with a single level about mfe

[lpfgarcia]

I made some tests using iris and an additional attribute with a single value. The first test added a numeric attribute and the second test added a factor attribute:

library(mfe)

data(iris)
iris$x = 1
metafeatures(Species~. , iris)

data(iris)
iris$x = factor(1)
metafeatures(Species~. , iris)

The first test returns an (expected) warning in the cov function. The second test returns an error caused by the single level attribute in the model.matrix (in the binarize function).

I checked how the other packages deal with that. While the rpart and nnet remove the attribute, the e1071 and kknn return the same error.

data(iris)
iris$x = factor(1)

library(e1071)
svm(Species~., iris)

library(kknn)
kknn(Species~., iris)

library(rpart)
rpart(Species~., iris)

I wish close the issue