I do not think so since there is a temporal shift between train and test, but it can be a good base for aggregation features. and since some users are only in the test set a proper cv startegy will be by holding some users to validate on so you get to mimic the train test split on predicting on unseen users.
I do not think so since there is a temporal shift between train and test, but it can be a good base for aggregation features. and since some users are only in the test set a proper cv startegy will be by holding some users to validate on so you get to mimic the train test split on predicting on unseen users.
Thanks Satti for this insightful idea.