Get user neighborhood from additional data in collaborative filtering

Question

I wanted to do recommendation based on multiple datasets like in Utilizing multiple, weighed data models for a Mahout recommender

But my problem is that additional data sets does not translate well into primary items. My domain is specific but imagine this as problem of recommending movies also based on music taste.

If I treat movies and music in the same manner I will get music in the results also.

I would like to use other data sets only to get better neighborhood. Is this possible with Mahout or Myrrix?

Answer 1

If you treat them in the same manner, you could use the Rescorer class to filter out the music results. Myrrix also has access to Rescorer(s)

For instance, this Rescorer example from the Mahout in Action book filters out from the recommendations the books that are not in stock. You would do something similar where you would keep (somehow) track of which item ids are music and then use that information in the Rescorer to filter them out so they don't get recommended.