Warning message: In `[<-.factor`(`*tmp*`, y == "benign", value = 0) : invalid factor level, NA generated

Question

I am using BreastCancerData from the R dataset to run a machine learning algorithm. However, when I try to convert the label "benign" to 0 and "malignant" to 1, it throws a warning message, and labels are not converted. The warning message is: Warning message: In `[<-.factor`(`*tmp*`, y == "benign", value = 0) :   invalid factor level, NA generated

How can I fix this warning message?

library(xgboost)

library(mlbench)

# Wisconsin Breast Cancer Database

data("BreastCancer")

y <- BreastCancer$Class

y[y=="benign"] = 0

y[y=="malignant"] = 1

Answer 1

As it is clear from the warning message that the class of the variable "y" is "factor" and you are comparing it with "character ". You need to un-factorize a string using the as.character() function, and, then, you can compare it with character.

How to check the class of a variable:

> library(xgboost)
> library(mlbench)
> data("BreastCancer")
> y <- BreastCancer$Class
> class(y)
[1] "factor"
> y <- as.character(BreastCancer$Class)
> class(y)
[1] "character"

So, make the highlighted changes in your code and it should work.

library(xgboost)
library(mlbench)

# Wisconsin Breast Cancer Database
data("BreastCancer")

# generate data labels
y <- as.character(BreastCancer$Class)
y[y=="benign"] = 0
y[y=="malignant"] = 1
y <- as.integer(y)