While performing the SMOTE for balancing the class data, what should be the proportion of both class? For instance, if we have 100 instances, what (%) should be the Yes class and what should be the No class?

  • 1
    $\begingroup$ There is no rule of thumb to determine that parameter. Only thing to remember is that SMOTE creates synthetic observations so over use might be detrimental for the model. I would suggest trying various combinations and checking which works for you better. $\endgroup$ Jul 3, 2019 at 0:13
  • $\begingroup$ @AmanMathur, thanks a lot $\endgroup$
    – Khan
    Jul 3, 2019 at 8:06

1 Answer 1


As I have mentioned in the comment, there is not rule of thumb. One has to do a lot experimentation.

Based on how much data you have, bringing minority class to be be equal (50-50) to the majority class can be misleading the model.

You can also try ADASYN during your experimentation.

ADASYN is a improved version of SMOTE. What it does is same as SMOTE just with a minor improvement. After creating those sample it adds a random small values to the points thus making it more realistic. In other words instead of all the sample being linearly correlated to the parent they have a little more variance in them i.e they are bit scattered.

You can read about the implementation here


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.