How could we express why we just drop the bids with missing country entry?

How will you calculate the proportion of bots for each country and the proportion of bots per device for test data? Because in test data there is no outcome column.

This is what you used for train data, like for test data how will create this feature to fed to model for prediction?

proportion of bots for each country

pbots_country = bid[bid['outcome'] == 1].groupby(
'country').size()/bid.groupby('country').size()

proportion of bots per device

pbots_device = bid[bid['outcome'] == 1].groupby(
'device').size()/bid.groupby('device').size()