Github user MLnick commented on the pull request:
https://github.com/apache/spark/pull/12922#issuecomment-217967241
The evaluator currently supports area under roc and pr. Those metrics are
across varying threshold levels. I don't see that they can actually be used
to pick a threshold, unless I'm missing something.
Accuracy could just be computed from the raw predicted and actual values as
#correct / total, but I guess that would be at the default threshold.
Unless the threshold is a param (which offhand I don't think it is), in
which case it could be cross-validated using, say, accuracy as a metric.
On Mon, 9 May 2016 at 18:11, Sean Owen <[email protected]> wrote:
> Hm, yeah I also agree that I have not seen an 'accuracy curve'. Maybe we
> don't add that for now after all.
>
> Nick how does the evaluator then work with respect to picking a threshold?
>
> â
> You are receiving this because you were mentioned.
> Reply to this email directly or view it on GitHub
> <https://github.com/apache/spark/pull/12922#issuecomment-217910223>
>
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
