You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Running a set of leave-one-in models, I'm getting different numbers of entities with predictions and even different numbers of labeled examples in the test set depending on which feature set is being included. It seems to be the case that entities without any record in the included feature table are being excluded entirely, even if they have labels in the test set.
The text was updated successfully, but these errors were encountered:
If merged, this commit will:
- change the logic for creating entity-date sets so that for train sets, it filters the labels against and entities table and for test sets, it creates a cross join of all entities in the entities table and the dates for the matrix [Resolves#69]
- creates entity date tables as temporary tables [Resolves#60], [Resolves#51]
- reformats entity date table names so that uuid is last [Resolves#50]
Running a set of leave-one-in models, I'm getting different numbers of entities with predictions and even different numbers of labeled examples in the test set depending on which feature set is being included. It seems to be the case that entities without any record in the included feature table are being excluded entirely, even if they have labels in the test set.
The text was updated successfully, but these errors were encountered: