Do we have the balanced error rate (BER)

DocMusher

Dear RM friends,
Perhaps I am unable to find this or it is not available or (a useless metric), but does RM provides the BER as a balanced metric that equally weights errors in Sensitivity and Specificity. 
All feedback is appreciated.
Sven

sgenzer

hi @DocMusher hmm never heard of it. Doesn't seem to hard to calculate, though.

IngoRM

Nope, this is not supported by any of the standard performance operators.  But since we deliver SEN and SPE it is indeed relatively easy to derive this as Scott has mentioned.  I have to admit that I have only encountered this measurement a handful of times in 20 years now so this may serve as a little comment about the usefulness

varunm1

Thats the trick part with research articles, some times people tend to be different and use some uncommon performance metrics. I used to search google when encountered with unheard metric while reviewing papers.There are also some styles which researchers follow, for example, in all my publications, I report kappa metric for performance, this is something I follow personally.

sgenzer

hi @DocMusher hmm never heard of it. Doesn't seem to hard to calculate, though.

IngoRM

Nope, this is not supported by any of the standard performance operators.  But since we deliver SEN and SPE it is indeed relatively easy to derive this as Scott has mentioned.  I have to admit that I have only encountered this measurement a handful of times in 20 years now so this may serve as a little comment about the usefulness

DocMusher

I just read it in this recent paper where it was used without further explanation but as THE performance feature? For what is worth. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0226962

MartinLiebig

@DocMusher,
if you need it, i'll write the operator for you. Maybe on Tuesday if we get the time. #RealHackathon.

Best,
Martin

DocMusher

So kind dear Martin, I just try to understand why it received so much attention in that PLOSOne paper while I think they could have used other features. I already responded on the PLOS website because I think they did not take care of te contamination examples published by Ingo a while ago (https://rapidminer.com/resource/correct-model-validation/)
Interested in your recent activities!!
Sven 

varunm1

Thats the trick part with research articles, some times people tend to be different and use some uncommon performance metrics. I used to search google when encountered with unheard metric while reviewing papers.There are also some styles which researchers follow, for example, in all my publications, I report kappa metric for performance, this is something I follow personally.

DocMusher

Thank you for placing this in a perspective. 
Sven