0

I have a data set with very imbalanced groups (about 3% have an effect, while 97% don't). I experimented with logistic regression and was convinced that I can get to a very good accuracy of prediction without matching.

In which cases is matching necessary, then?

Omry Atia
  • 597
  • 3
  • 11
  • Check https://stats.stackexchange.com/questions/283170/when-is-unbalanced-data-really-a-problem-in-machine-learning – Tim Mar 28 '19 at 10:26
  • Accuracy is probably one of the least convincing classification metrics when we are dealing with unbalanaced data. – usεr11852 Apr 01 '19 at 22:37

0 Answers0