I was wondering if a part from the accuracy it would be possible to have also the balanced accuracy in the Scorer node. Balanced accuracy is defined as:
Balanced accuracy = (sensitivity+specificity)/2
I find this measure very useful and I think it could represent a little improvement for the scorer node.
Thank you in advance,
That looks fairly easy for 2-class classification problems, and something that we can probably do with a math formula node right after the regular scorer.
Things get more complicated in multi-class problems and I'm not sure what the best approach would be there. It seems like this is a topic of active research but I am not at all an expert on the subject. I did find an interesting paper on this online with some sample R code that might be useful along with our R integration.
Now I understand why you didn't include that in the regular scorer node.
Thanks for the clarification.