Summary of - SVD-initialised K-means clustering for collaborative filtering recommender systems
Document Type
Article
Abstract
Study Background: The research presented titled “SVD-initialised K-means clustering for collaborative filtering recommender systems” by Dr Murchhana Tripathy, Dr Santilata Champati, and Dr Srikanata Patnaik presents a novel method of K-means initialization in the context of collaborative filtering recommender system. SVD (Singular value decomposition ) discovers natural clusters in a rating dataset, and thus, the number of clusters discovered by SVD can be considered as the initial K value for the K-means algorithm.
Research Goals and Hypotheses: The primary goal is to automate the K-means cluster initialization process.
Methodological Approach:
- Do the SVD of the given rating matrix X
- Determine user groups by multiplying the left singular vector matrix and the matrix containing the singular values. Let the number of user groups found be K.
- Use K found in step-2 to perform K-means of X
- Check if the user groups found by SVD is similar to the user clusters found by Kmeans. (A user group is considered similar to a user cluster when the same users that constituted a group in SVD also constitute a cluster in K-means. If both are similar, then SVD can be used to initialise K-means. Also, check if the clusters are stable.)
- Check the rank of matrix X. If K is less than the rank then slowly start increasing K up to the rank of the matrix and do K-means for each K value. Check if clustering is stable. If clustering is stable, then this K can be considered as the number of clusters and if not then go back to the previous K value where stable clusters were formed and consider that K as the final value for the number of clusters to perform K-means.
Results and Discoveries
Theoretically proved that SVD is a suitable method to solve and automate the K-means initialization problem.
Citation to the base paper - Tripathy, M., Champati, S., & Patnaik, S. (2022). SVD-initialised K-means clustering for collaborative filtering recommender systems. International Journal of Management and Decision Making, 21(1), 71-91.
Publication Date- December, 2021
Recommended Citation- Tripathy, M., Champati, S., & Patnaik, S. (2022). SVD-initialised K-means clustering for collaborative filtering recommender systems. International Journal of Management and Decision Making, 21(1), 71-91.
Publication Date
2022
Recommended Citation
Tripathy, Murchhana; Champati, Santilata; and Patnaik, Srikanata, "Summary of - SVD-initialised K-means clustering for collaborative filtering recommender systems" (2022). Open Access archive. 9226.
https://impressions.manipal.edu/open-access-archive/9226
