Details
-
New Feature
-
Status: Closed
-
Minor
-
Resolution: Won't Fix
-
None
-
None
-
None
-
None
Description
Q: Is there an EnsembleRecommender or CompoundRecommender that takes input
from other recommender algorithms and combine them to generate better
results?
Ted Dunning:
There isn't really any such thing although the SGD models are easy to glue
together in this way.
There is a guy named Praneet at UCI who is doing some feature sharding work
that might relate to what you are doing. His email is
praneetmhatre@gmail.com
Sean Owen:
There isn't. For the recommenders that work by computing an estimated
preference value for items, I suppose you could average their
estimates and rank by that.
More crudely, you could stitch together the recommendations of
recommender 1 and 2 by taking the top 10 amongst each of their top
recommendations – averaging estimates where an item appears in both
lists. That's much less work for you; it's not quite as "accurate".
Danny Bickson:
In terms of papers about ensemble methods/blending I suggest looking at the
BigChaos Netflix paper:
http://www.*netflixprize*.com/assets/*GrandPrize2009*_BPC_*BigChaos*.pdf
See section 7.