There are two types of recommender algorithms:
- Content-based filtering
- Collaborative filtering
In content-based filtering, each user is assumed to operate independently. Recommendations to the user are based on keywords, which represent an item.
In collaborative filtering, an item is associated with the ratings users have provided about this item collectively. The basic premise of collaborative filtering is this, if two users like a subset of items, there is a good chance they will like the rest of the items in the set too.
The most interesting characteristic of collaborative filtering is that it is neutral to the type of item being rated.