What is the exact definition for "imbalanced datasets" or "imbalanced learning"? Some article states it as,

In such problems, almost all the instances are labeled as one class, while far fewer instances are labeled as the other class, usually the more important class.

While some other article, like Neelam Rout (2018) says,

The ratio between the majority and minority classes may be 100:1, 1000:1 and 10000:1; in short, the instances of majority class outnumber the amount of minority class instances.

I have a dataset with 85% labeled positive and the others negative. Is it an imbalanced dataset?

  • $\begingroup$ Despite the existence of this tags, this question is better suited for Data Science SE or Stats SE. $\endgroup$ – Git Gud Dec 30 '18 at 13:01
  • $\begingroup$ Anyway, I would consider a ratio of approximately 5.67 to 1 to be imbalanced. This is highly subjective, there isn't a well defined threshold. Plus, it depends on what you're doing. If you're just analyzing, classifying as imbalanced is just an adjective, it doesn't matter. If you're going to train on this dataset, then it depends on what technique you're using and what your goal is. $\endgroup$ – Git Gud Dec 30 '18 at 13:08
  • $\begingroup$ I meant "these tags" in this comment. $\endgroup$ – Git Gud Dec 30 '18 at 13:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.