Hi everyone,
How can I combine content based recommender system with K-means clustering?
The items were clustered by using K-means algorithm.
What's the best way to apply a content-based recommendation ?
Thank you.
The first thing to do is to cluster your dataset based on some features in them that are appropriate for the type of recommendation you want to generate by the use of k-means clustering algorithm. Then, the next step is to apply content-based algorithm to generate recommendations based on the specific cluster that is similar to the user that needs a recommendation. The link below is a practical example of the application and implementation of K-means with content-based filtering using a movie dataset.
https://rpubs.com/tim-dim/recommender
Best!
Hi Folasade,
Thank you for the aswer, but the link you posted don't work.
A have dataset which contains Items, Users, and rating(User, Item).
After creating clusters, I want to learn user profiles from this clusters. Is it a good idea ?
Thnaks
Check the link again, it works.
Your question is not clear. Did you cluster items or users? The essence of clustering is to alleviate sparsity of data within the user-item matrix. After clustering your data, you could go ahead and make predictions based on the cluster that is similar to the user that wants recommendations with an appropriate variation of CF. I don't see the need to use CBF since there are no feature vectors in your matrix.
Best!
I cluster items.
I have feature vectors in my context. Every item is represented as a vector.
Thanks.
You may have to cluster users according to their preferences i.e. according to how similar the items they purchased are. Such that if a user purchases a new item same could be recommended to other users in the cluster that he/she belongs. Also, you can perform market-basket analysis on the items using association rule mining such that you know items that have high co-occurrence in the transaction set. That would help you be able to recommend associated items to items that users have shown interest in.If you wish to apply k-means algorithm in clustering, I have applied it in customer segmentation. See my paper below(The MATLAB code for the k-means algorithm is right there in the appendix). Am sure it may help. https://www.researchgate.net/publication/282862569_Application_of_K-Means_Algorithm_for_Efficient_Customer_Segmentation_A_Strategy_for_Targeted_Customer_Services
Article Application of K-Means Algorithm for Efficient Customer Segm...
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling