side information about users

11 views

Skip to first unread message

Andrew Ziem

unread,

Jun 13, 2019, 7:10:27 PM6/13/19

to LensKit Recommender Toolkit Development and Support

Is there a way of adding side information to Lenskit models? By side information, I mean non-rating user features such as user's demographics that is helpful to predict ratings. It seems this would be especially helpful in the cold-start situation.

In the study "Tears or Fears? Comparing Gender Stereotypes about Movie Preferences to Actual Preferences" (2017), authors concluded men and women have distinctive preferences for movie genres.

Best regards,

Andrew

Michael Ekstrand

unread,

Jun 13, 2019, 7:13:50 PM6/13/19

to Andrew Ziem, LensKit Recommender Toolkit Development and Support

Andrew,

None of LensKit's algorithms make use of such data, but you can implement an algorithm that takes additional data tables to its 'fit' method (in LensKit for Python). Right now there are a few oddities in how that relates to helper algorithms, like the Top-N wrapper; we would like to address these to make it work smoothly (e.g. have it pass on additional parameters untouched), but would like to have an example of such an algorithm to make sure we do it in an ergonomic fashion.

- Michael

--
You received this message because you are subscribed to the Google Groups "LensKit Recommender Toolkit Development and Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lenskit-recsy...@googlegroups.com.
To post to this group, send email to lenskit...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lenskit-recsys/71e0a21e-d3d3-4545-9900-7ed0df751c1e%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Michael D. Ekstrand — michael...@boisestate.edu — https://md.ekstrandom.net
Assistant Professor, Dept. of Computer Science, Boise State University

People and Information Research Team (PIReT) — http://coen.boisestate.edu/piret/