Title: Collaborative filtering by sequential user-item co-cluster extraction from rectangular relational data

Authors: Katsuhiro Honda, Akira Notsu, Hidetomo Ichihashi

Addresses: Graduate School of Engineering, Osaka Prefecture University, 1-1 Gakuen-cho, Sakai, Osaka, 599-8531, Japan. ' Graduate School of Engineering, Osaka Prefecture University, 1-1 Gakuen-cho, Sakai, Osaka, 599-8531, Japan. ' Graduate School of Engineering, Osaka Prefecture University, 1-1 Gakuen-cho, Sakai, Osaka, 599-8531, Japan

Abstract: Collaborative filtering is a basic technique for tackling information overloads and is composed of task of relating a promising item to an active user. In this paper, a new approach to user-item co-cluster extraction from rectangular relational data is proposed based on the structural balancing concept, and the clustering method is applied to collaborative filtering tasks. In the process, user-item rectangular relational matrix given in an alternative process of |liking or not| is first transformed into a square adjacency matrix and then co-clusters are sequentially extracted by using a weighted aggregation criterion. In a numerical experiment, the proposed collaborative filtering model is applied to a purchase history data set in order to demonstrate the recommendation ability of the model.

Keywords: collaborative filtering; fuzzy clustering; relational data; rectangular relational data; information overload; structural balancing; user-item co-cluster extraction; co-clusters; purchase history; recommendation ability; modelling.

DOI: 10.1504/IJKESDP.2010.037493

International Journal of Knowledge Engineering and Soft Data Paradigms, 2010 Vol.2 No.4, pp.312 - 327

Published online: 14 Dec 2010 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article