How do you deal with unbalanced classification problem?

Question

Anonymous · Accepted Answer

Sampling technique like oversampling and downsampling. Models that adapt to unbalanced data like Ada-boosting. Or use anomaly detection methods.

arman · Answer

The basic idea is sampling (oversampling like resampling) which is not a good option when dealing with huge imbalance.  Another option is to set weights for class labels (larger weights for classes with fewer samples). A really good, but tough and many times not feasible to implement, is to do oversampling by using generative networks (i.e. GANs).

Anonymous · Answer

over/undersampling, GANs (I believe they require lots of data), and class weights (sometimes refer to class weights)

Capital One

Capital One interview question

Interview Answers

Want the inside scoop on your own company?

Bowls

Followed companies

Job searches