On Collaborative Filtering with Possibilistic Clustering for Spherical Data Based on Tsallis Entropy

This paper proposes a collaborative filtering (CF) method using possibilistic clustering for spherical data based on Tsallis entropy. This study was motivated by a previous work, which showed that adopting fuzzy clustering for spherical data in CF tasks provided better recommendation accuracy than fuzzy clustering for categorical-multivariate data. Moreover, possibilistic clustering algorithms are naturally more robust to noise than fuzzy clustering. The results of experiments conducted on an artificial dataset and one real dataset indicate that the proposed method is better than the conventional methods in terms of recommendation accuracy.