Thanks for releasing the Machine Learning content. I recently finished it and I had a question about lesson 12 (1:40) the rossman competition.
You talked about using days since last holiday and days to next holiday as new features for each row in the dataset. You also talked about doing the same for the promo feature. It sounds valid to do this for holidays as they are fixed and know ahead of time however I didn’t think it was valid to do this for promo because the decision to run a promo may depend on past performance which is not known ahead of time. For example, the decision to run a promo may depend on past sales since previous promo and we may not have a schedule of future promo’s in the real world. Do you think my thought process is valid?
Once again, thanks for releasing this course. I definitely learnt some new techniques I have not seen in other courses.