Sep 7, 2018 · 1 min read
Thank you for a great post.
I am curious about your interpretation of the encoded categorical features generated by RFE. For example, RFE calculated that Wednesday and Friday were among the most important features. It is my impression that sklearn feature selection tools do not offer categorical feature support (https://github.com/scikit-learn/scikit-learn/issues/8480). In other words, RFE doesn’t know that day_of_week_wed is an encoded categorical feature. That said, how do you evaluate whether day_of_week is important or only selected days (i.e. Wednesday and Friday) are informative?