2013/7/2 Gene Kogan <[email protected]>: > That is a good question. I'm not sure what the recommended approach is. I > suppose I can use the binary classification accuracy as my evaluation and > trust the continuos outputs from that are representative. I don't know if > this is good practice or not, but I can't think of another approach since I > don't have continuous ground truth.
I think log loss is the common way of evaluating probabilities with discrete ground truth. I have a pull request pending to add that to scikit-learn [1]. You might be able to use some of the code. [1] https://github.com/scikit-learn/scikit-learn/pull/2013 -- Lars Buitinck Scientific programmer, ILPS University of Amsterdam ------------------------------------------------------------------------------ This SF.net email is sponsored by Windows: Build for Windows Store. http://p.sf.net/sfu/windows-dev2dev _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
