Spherical Rough C-Means Clustering and Its Application to Collaborative Filtering

Content recommendation systems, seen on e-commerce sites like Amazon and video streaming services like YouTube and Netflix, utilize collaborative filtering (CF). Improving content recommendations enhances usability, purchase rates, viewership, and corporate profits. Clustering-based CF extracts clusters of users with similar preference patterns and recommends highly pre-ferred content within the cluster to which the user belongs. Due to the inherent ambiguity and uncertainty in user preference data, traditional hard C-means (HCM; k-means) clustering is inadequate. This study focuses on rough clustering based on rough set theory to handle the uncertainty inherent in data. We propose spherical rough C-means (SRCM) clustering, which clusters ob-jects on a unit hypersphere using cosine similarity, addressing un-certainty in cluster memberships. As an application, we present SRCM-CF, a collaborative filtering method based on SRCM. Furthermore, we verified the recommendation performance of the proposed method using a real-world dataset.