0

My understanding when it comes to unbalanced datasets is that we can randomly sample from the dominant class.

What are some ways to deal with unbalanced data when we have time series data and the goal is to either build a boosting tree or neural network?

Thanks!

confused
  • 2,453
  • 6
  • 26
  • [Are unbalanced datasets problematic, and (how) does oversampling (purport to) help?](https://stats.stackexchange.com/q/357466/1352) – Stephan Kolassa Mar 11 '20 at 08:10
  • Yes but you can't really randomly sample time series data – confused Mar 11 '20 at 08:47
  • The linked thread and others make it clear that (over-)sampling is not a solution, because there is no problem in the first place. Simply model a binary time series, e.g., using a logit link or similar. Forecast a probability. This will be low, but that's correct. Then think about what decisions you want to take based on this probability prediction. – Stephan Kolassa Mar 11 '20 at 08:53

0 Answers0