Comparison between user-based and item-based collaborative filtering algorithms

First of all, there are two types of collaborative filtering algorithms: user-based collaborative filtering algorithm (usercf) and item-based collaborative filtering algorithm (itemcf ).

The user-based collaborative filtering algorithm consists of two steps:

1) find a set of users with similar interests to the target user

2) Find the items that the users in this set like and the target users have never heard of and recommend them to the target users.

The item-based collaborative filtering algorithm consists of two steps:

1) calculate similarity between items.

2) generate a recommendation list for users based on item similarity and historical user behavior.

We can see that usercf is a hot topic in the interest group where users are recommended, and focuses more on socialization. itemcf recommends similar items based on users' historical behaviors and focuses more on personalization. Therefore, usercf is generally used in news websites, such as Digg, while itemcf is used in other non-news websites, such as Amazon and Hulu.

In news websites, users' interests and hobbies are often coarse-grained. Few users say that they only watch the news of a specific topic, and sometimes there is not news on a certain topic every day. Personalized news recommendations are more popular news. Popularity and timeliness are the focus of personalized news recommendations. personalization is a supplement. Therefore, usercf recommends news to users who share the same interests and interests, in this way, both hotspot and timeliness are ensured, while personalization is taken into account. Another reason is that technically, as an item, news is updated very quickly and there will be new news in real time. If itemcf is used, we need to maintain a table of similarity between items. In actual industry, this table is updated every day, which is unacceptable in the News Field.

However, itemcf can play a better role in books, e-commerce, and movie websites. In these websites, users' interests and interests are generally fixed and more delicate than those of news websites. In these websites, personalized recommendations are generally used to recommend users related items in their own fields. In addition, the update speed of the number of items on these websites is not fast, and one update per day is acceptable. In addition, the number of users in these websites is often far greater than the number of items. From the storage point of view, usercf consumes more space and complexity. In addition, itemcf can easily provide recommendations, this increases user trust in the recommendation system, so it is more suitable for these websites.