UserCF principle: UserCF recommended that the user and his articles have common interests likes to user
ItemCF principle: ItemCF similar to those recommended by his previous favorite items and items to the user
UserCF recommended more social, reflecting the interests of small groups of items where the user popularity ; and ItemCF recommended more personalized , reflecting the user's own interest in heritage
UserCF suitable News Recommended reason:
- Popularity and timeliness are personalized news recommendation of focus, and personalize with respect to these two somewhat minor
- UserCF need to maintain a similar user interest table, and ItemCF need to maintain a similar list items, news articles recommended system update speed is very fast, so if ItemCF then, the similarity table items need to be updated soon, which It is difficult to achieve
ItemCF suitable for reasons of books, movies and e-commerce websites:
- The user's interest is fixed and relatively long-lasting
- Knowledge of these systems require less user popularity to help them judge the quality of an item, but by their own areas known to judge the quality of their own articles
UserCF of Applications:
- Fewer users occasion, if a lot of users, user computing similarity matrix is costly (news website)
- Timeliness strong, user personalization less obvious areas of interest
- So that users do not need to give convincing explanation recommendation
ItemCF of Applications:
- The number of items suitable for significantly less than the number of users of the occasion, such as a lot of items (sites), calculate the similarity matrix is costly items
- Nagao items rich, strong user demand for personalized areas
- Users need to use the historical behavior of the user do recommend explained, can make users more convincing