Abstract
Recommender systems assist the e-commerce providers for services computing in aggregating user profiles and making suggestions tailored to user interests from large-scale data. This is mainly achieved by two primary schemes, i.e., memory-based collaborative filtering and model-based collaborative filtering. The former scheme predicts user interests over the entire large-scale data records and thus are less scalable. The latter scheme is often unsatisfactory in recommendation accuracy. In this paper, we propose Large-scale E-commerce Recommendation Using Smoothing and Fusion (CFSF) for e-commerce providers. CFSF is divided into an offline phase and an online phase. During the offline phase, CFSF creates a global item similarity matrix (GIS) and user clusters, where user ratings within each cluster is smoothed. In the online phase, when a recommendation needs to be made, CFSF dynamically constructs a locally-reduced item-user matrix for the active user item by selecting the top M similar items from GIS and top the K like-minded users from user clusters. Our empirical study shows that CFSF outperforms existing CF approaches in terms of recommendation accuracy and scalability.