Improving Collaborative Filtering by W-clustering

H. Chen, K. Furuse, N. Ohbo, and S. Nishihara (Japan)


Collaborative Filter, Information Searches, Knowledge Discovery, Digital Library, Cluster ing.


Collaborative filtering has been very successful in appli cations such as E-commerce and digital library. In this paper we propose a new approach for improving collab orative filtering by W-clustering (Voters and Voted-items clustering). That is, items are clustered and users are also clustered with respect to item clusters. Collaborative filter ing is then performed on cluster-cluster relationship which overcomes several problems which traditional methods are suffered from. More specifically, comparing to tradition methods, cluster-cluster collaborative filtering avoids the problem of sparsity and the so called cold start problem, and enhances the accuracy of finding neighbor users hence results in a more effective collaborative filtering. Experi ments on the EachMovie data set show favorable results.

Important Links:

Go Back