Since there might be daily effects, you might not predict the average click rate or conversion rate, but the area under the curve, which represents the likelihood of giving the positive example a score higher than a negative example, should be reasonably close.
To keep things simple, each model should either be an ensemble only taking the input of other models, or a base model taking many features, but not both
Перевод: Чтобы сохранить простоту, каждая модель должна быть либо ансамблем, принимая на вход результаты других модели, либо базовой моделью, использующей множество признаков, но не оба варианта сразу.
проверить корректность перевода
You also want to enforce properties on these ensemble models.
Identifiers of documents being retrieved and canonicalized queries do not provide much generalization, but align your ranking with your labels on head queries.
Что такое "canonicalized queries"?
Что такое head queries?
If you predict the probability that a document is spam and then have a cutoff on what is blocked, then the precision of what is allowed through matters