Increasing number of changesets flagged

[bkowshik]

For the last two days that gabbar has been live on osmcha-staging, it has flagged less than 30 changesets as problematic everyday. In the real-world, there could potentially be more changesets that are harmful. So:

• How can we increase the number of changesets flagged by gabbar?
• How can we bring accuracy down a little so that we bring number of false negatives down?

---

cc: @rodowi

[danstowell]

Hi - an osmcha user here. I see many dozens of "flagged by gabbar", and I'm only looking at a localish area - I don't think you want to increase the amount too much.

You say "bring accuracy down a little", but that can't be right. You want to increase the accuracy, in general. I guess you mean you want to change the threshold, i.e. change the balance of false-positive and false-negative. But I go back to my first point - do consider the risk of flooding osmcha!

[bkowshik]

:wave: @danstowell happy to meet a fellow osmcha user. 😃

gabbar is a super-experimental machine learning based changeset classifier to identify problematic edits on OpenStreetMap. We deployed to osmcha-django-production over the weekend, but it turned out to be super noisy and as a result we have brought it down. We will take the next couple of days to iterate on the classifier and bring back an improved version of gabbar to osmcha-django.

I guess you mean you want to change the threshold.

Thank you for clarifying. I meant changing the threshold of the classifier.

[bkowshik]

No next actions here. Closing.