|
USER PROFILING: COLLABORATIVE FILTERING AbstractCollaborative filtering is based on the assumption that “similar users have similar preferences”. In other words, by finding users that are similar to the active user and by examining their preferences, the recommender system can (i) predict the active user’s preferences for certain items and (ii) provide a ranked list of items which active user will most probably like. Collaborative filtering generally ignores the form and the content of the items and can therefore also be applied to non-textual items. Furthermore, collaborative filtering can detect relationships between items that have no content similarities but are linked implicitly through the groups of users accessing them. These groups (communities) are formed around a specific user profile.
[Edit] |
