There are three types of recommenders, knowledge based, content based, collaborative filtering. Within collaborative filtering, there two main branches, model Based Collaborative Filtering, neighborhood Based Collaborative Filtering.
1. Knowledge Based Recommendation
It is commonly used in luxury shopping as it is often built from a rank-based recommendation system with knowledge-based filters on user specific preference or settings. Often a rank-based algorithm is provided along with knowledge-based recommendations to bring the most popular items in particular categories to the user's attention. For example, we could write a function that returns items that are ranked by review rating, review count and review date while recommending to users. Used for luxury, buying cars or houses.
2. Content Based Recommendations
Content based recommendations are when we use information about the users or items to assist in our recommendations. In this recommendation technique, we use information that is known about the user or item to make recommendations. This method of making recommendations is particularly useful when we do not have a lot of user-item connections available in our dataset.
3. Collaborative Filtering
It is a method of making recommendations based only on the interactions between users and items. There are two main types of collaborative filtering which are user based and item based.
In item-based collaborative filtering, first you need to find the items that are most related to each other item (based on similar ratings). Then you can use the ratings of an individual on those similar items to understand if a user will like the new item.
In user-based collaborative filtering, users related to the user you would like to make recommendations for are used to create a recommendation. Using Netflix as an example, we want to recommend relevant movies for users to watch.
Note: It might be the case that content based and collaborative filtering based techniques come up with similar recommendations, but the methods by which data scientists approach these recommendations are very different. In collaborative filtering, you are using the connections of users and items (as you did before). In content based techniques, you are using information about the users and items, but not connections (hence the usefulness when you do not have a lot of internal data already available to use).
If you are interested in further readings on more complex recommenders, check out this article on how airbnb uses embedding to make its recommendations.