Unfortunately, neither MCC nor F-Measure are really suited in most cases of
imbalanced data, although they are way better than accuracy. Especially, with
F-Measure we got bad behavior due to changing class ratios in our data. If you
want to have an intuitive measure which does not use a shifting of the decision
boundary (as it hold for ROC curves) you might consider the weighted accuracy
(or balanced accuracy). For a (readable) review on this issue and a comparison
of metrics you might want to look at:
Sirko Straube, Mario Michael Krell (2014),
How to evaluate an agent’s behaviour to infrequent events? — Reliable
performance estimation insensitive to class distribution,
In Frontiers in Computational Neuroscience 8(43), doi:10.3389/fncom.2014.00043
------------------------------------------------------------------------------
Want fast and easy access to all the code in your enterprise? Index and
search up to 200,000 lines of code with a free copy of Black Duck
Code Sight - the same software that powers the world's largest code
search on Ohloh, the Black Duck Open Hub! Try it now.
http://p.sf.net/sfu/bds
_______________________________________________
Scikit-learn-general mailing list
Scikit-learn-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
