A graph based approach to extract a neighborhood customer community for collaborative filtering

Abstract

In E-commerce sites, recommendation systems are used to recommend products to their customers. Collaborative filtering (CF) is widely employed approach to recommend products. In the literature, researchers are making efforts to improve the scalability and online performance of CF. In this paper we propose a graph based approach to improve the performance of CF. We abstract a neighborhood community of a given customer through dense bipartite graph (DBG). Given a data set of customer preferences, a group of neighborhood customers for a given customer is extracted by extracting corresponding DBG. The experimental results on the MovieLens data set show that the recommendation made with the proposed approach matches closely with the recommendation of CF. The proposed approach possesses a potential to adopt to frequent changes in the product preference data set.