Cliques-based Data Smoothing Approach for Solving Data Sparsity in Collaborative Filtering

Abstract

Collaborative filtering (CF), as a personalized recommending technology, has been widely used in e-commerce and other many personalized recommender areas. However, it suffers from some problems, such as cold start problem, data sparsity and scalability, which reduce the recommendation accuracy and user experience. This paper aims to solve the data sparsity in CF. In the paper, cliques-based data smoothing approach is proposed to alleviate the data sparsity problem. First, users and items are divided into many cliques according to social network analysis (SNA) theory. Then, data smoothing proceeding is carried out to fill the missing ratings in user-item rating matrix based on the user and item cliques. Finally, the traditional user-based nearest neighbor recommendation algorithm is used to recommend items for users. The experiments show that the proposed approach can effectively improve the accuracy and performance on sparse data.