error in balancing with SMOTE

Asked

Viewed 264 times

1

Hello, I have a somewhat unbalanced dataset and wanted to do some testing with smote, but I’m getting an error:

library(DMwR)

treinoSmote <- SMOTE(TARGET ~ .,m,k=5, perc.over = 100, perc.under = 200)
Error in factor(newCases[, a], levels = 1:nlevels(data[, a]), labels = levels(data[,  : 
invalid 'labels'; length 0 should be 1 or 2

my TARGET is already a factor, I left it with values 1 and 0, with S and N (YES and NO) etc, always gives that error.

my dataset is composed of integer Features, factor and Numeric. There are about 20 at the moment.

The only things I see on the Internet say it must be factor and stuff, but it’s over!

I did the test that you have in the documentation of the SMOTE itself with the iris dataset and it works normal. I checked the type of the Feature and is as a factor as well. I don’t understand why you’re making this mistake.

data(iris)
data <- iris[,c(1,2,5)]
data$Species <- factor(ifelse(data$Species == "setosa", "rare", "common"))
table(data$Species)
common   rare 
100     50 

newData <- SMOTE(Species ~ ., data, perc.over = 600, perc.under = 100)
table(newData$Species)
common   rare 
300    350
r

  • It would be interesting to provide your database (or a part of it) to reproduce in the same way. Use the command dput

  • so serve? https://anotepad.com/notes/ind9p6 rsrs did not find place to attach or something like that

  • It is also worth saying which package used, since the function SMOTE is not a base R function. In these cases start at all times the question with the package loading, library(DMwR).

  • sorry, I had not seen that I had not put, I already edited there

  • pqp found the error. It had a Feature that had not converted correctly. Sorry the disorder. the F8 should be factor and was Character

1 answer

1

i had left a Feature with the type Character, had not converted to factor. It was not in the target Feature, it was in the middle.

I just used the a.factor and managed to rotate normally

m$F8 <- as.factor(m$F8)

Browser other questions tagged r

You are not signed in. Login or sign up in order to post.