Collaborative Filtering with q-Divergence-Based Relational Fuzzy c-Means Clustering

Abstract

This paper presents a collaborative filtering (CF) algorithm for a recommendation system motivated by the need for higher recommendation accuracy. Cluster analysis captures information from a group of users such that users within a given cluster are more similar to each other than to those in other clusters. Therefore, clustering helps to detect similar users. However, inconsistent similarity measures have been applied during the clustering and prediction stages in the literature. Hence, this study resolves such discrepancies through the proposed CF algorithm, which uses fuzzy clustering for relational data such that a common similarity measure is applied to both the clustering and prediction stages. Experiments were conducted with ten datasets based on an artificial dataset and 40 datasets based on eight real datasets to demonstrate that the proposed algorithm could achieve a higher CF accuracy than conventional methods.