Joining Items Clustering and Users Clustering for Evidential Collaborative Filtering

Abstract

Recommender Systems (RSs) are supporting users to cope with the flood of information. Collaborative Filtering (CF) is one of the most well-known approaches that have achieved a widespread success in RSs. It consists in picking out the most similar users or the most similar items to provide recommendations. Clustering techniques can be adopted in CF for grouping these similar users or items into some clusters. Nevertheless, the uncertainty comprised throughout the clusters assignments as well as the final predictions should also be considered. Therefore, in this paper, we propose a CF recommendation approach that joins both users clustering strategy and items clustering strategy using the belief function theory. In our approach, we carry out an evidential clustering process to cluster both users and items based on past preferences and predictions are then performed accordingly. Joining users clustering and items clustering improves the scalability and the performance of the traditional neighborhood-based CF under an evidential framework.