Basic Consideration of Collaborative Filtering Based on Rough C-Means Clustering
Seiki Ubukata, Shu Takahashi, Akira Notsu, Katsuhiro Honda · 2020
Collaborative filtering (CF) is a technique for realizing recommender systems found in e-commerce sites and video streaming services. Appropriate content recommendations to individual users will improve usability, purchase rates, video ratings, and corporate profits. Clustering is a technique for automatically classifying and summarizing the data by extracting clusters composed of similar objects. Clustering-based CF extracts clusters composed of users with similar interests and preferences, and recommends contents with high preference degree within the cluster. Rough C-means (RCM)-type methods can extract overlapped clusters dealing with the certainty, possibility, and uncertainty of belonging of object to clusters. In this study, we propose two-types of RCM-based CF, namely, GRCM-based CF (GRCM-CF) and$\pi$GRCM-based CF ($\pi$GRCM-CF). Furthermore, we verified the recommendation performance of the proposed methods, GRCM-CF and$\pi$GRCM-CF, through numerical experiments using two real-world datasets, namely, NEEDS-SCAN/PANEL dataset and MovieLens dataset.