4
Collaborative Filtering
˘Effective when untainted data is available
˘Typically have to deal with sparse data
lUsers will only vote over a subset of all items they’ve
seen
˘Data:
lExplicit: recommendations, reviews, ratings
lImplicit: query, browser, past purchases, session
logs
˘Approaches
lModel based – derive a user model and use for prediction
lMemory based – use entire database
˘Functions
lPredict – predict ranking for an item
lRecommend – produce ordered
list of items of interest to the user.
Why are these two considered distinct?