21.7 Applying Group Recommendation to Individual Users
So, what if you are developing an application that recommends to a single user? Group recommendation techniques can be useful in three ways: (1) to aggregate multiple criteria, (2) to solve the so-called cold-start problem, (3) to take into account opinions of others. Chapter 22 also discusses how aggregation may be needed when recommending to individuals, and covers several specific aggregation functions.
21.7.1 Multiple Criteria
Sometimes it is difficult to give recommendations because the problem is multidimensional: multiple criteria play a role. For instance, in a news recommender system, a user may have a preference for location (being more interested in stories close to home, or related to their favourite holiday place). The user may also prefer more recent news, and have topical preferences (e.g. preferring news about politics to news about sport). The recommender system may end up with a situation like in Table 21.5, where different news story rate differently on the criteria. Which news stories should it now recommend?