1

Is it possible to oversample the dataset without replacement? For RandomUnderSampling, there exist a boolean hyperparameter [replacement]; But, this hyperparameter doesn't exist in RandomOverSampling

Looking at RandomOverSampling Docs:

Object to over-sample the minority class(es) by picking samples at random with replacement. The bootstrap can be generated in a smoothed manner.

Any suggestions or alternative approaches to follow?

Ahmad hassan
  • 1,039
  • 7
  • 13
  • This is just a warning and does not actually produce any error right? – yudhiesh Dec 09 '20 at 12:10
  • Yes, It is a warning. But, "Use at your own risk", and for machineLearning models may fail with producing inconsistent results. – Ahmad hassan Dec 09 '20 at 12:12
  • Ok then just delete sklearn using this `pip uninstall scikit-learn` then install the version you need using `pip install scikit-learn=0.19.1`. – yudhiesh Dec 09 '20 at 12:20

0 Answers0