AKhatun_WMF added a comment.
In T288262#7628599 <https://phabricator.wikimedia.org/T288262#7628599>, @MPhamWMF wrote: > @AKhatun_WMF , sorry, it's been a while since I wrote this, but I think what I meant when I wrote the question about "optimal separation" is given some distribution of ORES scores (e.g. a normal distribution), is it clear what the threshold is for what qualifies as a "high" vs "low" score: e.g. anything over .75 is a high score. But that's assuming the scores are continuous. I guess it's moot if they're binary (I don't actually know). > > If this isn't a sensible way of thinking about the issue, let me know if there's a better way. Ah, that I believe is already solved by the output of the model. Basically, we get probabilities for 5 classes (A to E) determining how good an item is, where A is the best and E is the worst. And then the score is calculated as `5*ProbabilityOfClassA + 4*ProbabilityOfClassB + 3*ProbabilityOfClassC + 2*ProbabilityOfClassD + 1*ProbabilityOfClassE`. But we can definitely define our own thresholds as well. The analysis is done here: Wikidata_Item_ORES_Score_Analysis <https://wikitech.wikimedia.org/wiki/User:AKhatun/Wikidata_Item_ORES_Score_Analysis> I will be doing a bit more to get the scores per subgraph and will add it here as well. TASK DETAIL https://phabricator.wikimedia.org/T288262 EMAIL PREFERENCES https://phabricator.wikimedia.org/settings/panel/emailpreferences/ To: AKhatun_WMF Cc: JAllemandou, dcausse, ACraze, Aklapper, AKhatun_WMF, Addshore, Manuel, MPhamWMF, Gethan, Simonmaignan, Invadibot, maantietaja, calbon, lmata, Anerka, CBogen, Akuckartz, Nandana, Namenlos314, Lahi, Gq86, Xinbenlv, Vacio, Lucas_Werkmeister_WMDE, GoranSMilovanovic, Fz-29, QZanden, EBjune, merbst, LawExplorer, elukey, _jensen, rosalieper, Mkdw, Scott_WUaS, Jonas, Xmlizer, notconfusing, jkroll, Wikidata-bugs, Jdouglas, aude, Tobias1984, Manybubbles, Alchimista, Mbch331
_______________________________________________ Wikidata-bugs mailing list -- wikidata-bugs@lists.wikimedia.org To unsubscribe send an email to wikidata-bugs-le...@lists.wikimedia.org