Page MenuHomePhabricator

Calculate expected reverts per day for multilingual revert risk model
Closed, ResolvedPublic

Description

As in T372280, please generate datasets for reverts per day at different thresholds for the multilingual revert risk model.

We don't know the thresholds yet, and will probably need to wait for T372747 datasets to make educated guesses.

Event Timeline

@Samwalton9-WMF Is this required for all wikis, or the largest N number of wikis or a specific set of wikis?

@Samwalton9-WMF Is this required for all wikis, or the largest N number of wikis or a specific set of wikis?

I think we can probably start with the list we published here, but also including Serbian, Latvian, and Punjabi Wikipedias, as they've all expressed some interest since the original data was published.

@Samwalton9-WMF I did an initial aggregation here. Depending on what we get from the testing process, we can adjust the thresholds and re-calculate.

KCVelaga_WMF changed the task status from Open to In Progress.Sep 19 2024, 2:44 PM
KCVelaga_WMF moved this task from Next 2 weeks to Doing on the Product-Analytics (Kanban) board.

Thank you! We'll revisit this data to put on our new testing page once we've figured out the rough initial thresholds.

@Samwalton9-WMF I did an initial aggregation here. Depending on what we get from the testing process, we can adjust the thresholds and re-calculate.

Could you re-run these for 0.95, 0.96, 0.97, 0.98, and 0.99?

Could you re-run these for 0.95, 0.96, 0.97, 0.98, and 0.99?

In the same notebook, I added 0.96 to the overall table and also a table at the end with just the given thresholds, and a CSV file of the same.