Using incremental clustering technique in collaborative filtering data update

Abstract

Collaborative filtering (CF) techniques are widely used by online shops in their recommender systems. It is well known that the nonnegative matrix factorization (NMF) based CF algorithms are popular and can provide reasonable product recommendations. However, the dimensions of the factor matrices in NMF need to be predetermined and updated when necessary. Moreover, data arrives in every second so the recommender systems must be capable of updating the fast growing data in a timely manner. In this paper, we propose an approach that incorporates incremental clustering technique into NMF based data update algorithm which can determine the dimensions of the factor matrices and update them automatically. The approach clusters users' and items' auxiliary information and uses them as constraints in NMF for data update. The cluster quantities are used as the dimensions of the factor matrices. With more data coming in, the incremental clustering algorithm determines whether to increase the number of clusters or merge the existing clusters. Experiments on three different datasets (MovieLens, Sushi and LibimSeTi) are conducted to examine the proposed approach. The results show that our approach can update the data quickly and provide encouraging prediction accuracy.