Technique for modelling when number of responders are less


Can you suggest modelling technique which work well when the number of responders is very small(Approximately 2-3% in the entire dataset). Linear/Logistic regression fails in such cases?

Thanks in advance.


@asinghan- I would suggest you to try Linear discriminant analysis it works very well when number of rows are very small.

Hope this helps!



I think what you mean is that there is a class imbalance in the dataset. Approximately 2-3% are responders while the rest 97-98% are non-responders. Is this the right understanding?

In this case, the following blog post will be of very helpful to you.