Another Pro for categorical summary encoding

[EmilHvitfeldt]

In https://feaz-book.com/categorical-summary, an extra pro will be that "further information can be numerically represented". In the past extra adding the quantile=0.9, helped us improve uncertainty estimates. So, representing the category using a few quantiles can provide extra juicy information.

Originally posted by https://github.com/cmougan in https://www.linkedin.com/feed/update/urn:li:activity:7279549399720935424?commentUrn=urn%3Ali%3Acomment%3A%28activity%3A7279549399720935424%2C7279556503391399936%29&dashCommentUrn=urn%3Ali%3Afsd_comment%3A%287279556503391399936%2Curn%3Ali%3Aactivity%3A7279549399720935424%29

[EmilHvitfeldt]

Also says

In summary encoder, even if its more computationally expensive, it's not so expensive (compared to deeply learning methods). Perhaps a rewording.