User-based collaborative filtering evaluates the similarity between users through the ratings of different users on items, and makes recommendations based on the similarity of users.
The core idea of the algorithm: In an online recommendation system, when user A needs a personalized recommendation, he can first find other users who have similar interests to him, and then recommend those items that the user likes but the user A has not heard of. A, This method is called user-based collaborative filtering algorithm.
==> It can be seen that this algorithm mainly includes two steps:
1. Find a set of users with similar interests to the target user - calculate the interest similarity between the two users
2. Find the items that the users in this collection like and that the target user has not heard of and recommend to the target user - find the item to recommend
Find similarity between users
1, Jaccard official
2. Pearson correlation coefficient
3. Euclidean distance
4. Cosine distance
Analysis of advantages and disadvantages of user-based collaborative filtering algorithm
Analysis of advantages:
First of all, it can classify users according to the similarity of their ratings of items through mutual assistance between users, and the obtained results are relatively accurate. Secondly, in the user-based collaborative filtering system, all users can benefit from the feedback and evaluation of neighbor users. As long as each user contributes to the system, the performance of the system will become more and more perfect. Finally, the user-based collaborative filtering system can easily tap the potential new interests of target users, that is, it can realize singularity discovery.
Disadvantage analysis:
1. Sparsity. A large-scale e-commerce recommendation system generally has a lot of items, and users may buy less than 1% of the items. The overlap between the items purchased by different users is low, so that the algorithm cannot find a user's neighbors, that is, preferences. similar users.
2. Cold start problem. When a new item appears for the first time, no user has evaluated it, and user-based collaborative filtering cannot predict and recommend it. Moreover, since there are few user evaluations in the early stage of new items, the accuracy of recommendation is not high.
3. Special user issues. A small number of users with special preferences will not benefit from the user-based collaborative filtering recommendation system, because the user-based collaborative filtering recommendation is based on the neighbor user data to obtain the recommendation for the target user, and its calculation implies a premise that each user is Must have relatively concentrated and fixed hobbies. However, the views of these special users are different from those of any group, and it is impossible to find neighbor users to make recommendations. Even after the initial stage of the system has passed, it is difficult for these special users to get more accurate recommendations from the collaborative filtering system.