COBA: A Credible and Co-clustering Filterbot for Cold-Start Recommendations

Collaborative Filtering (CF) assists Recommender Systems (RSs) in recommending products or services that they are likely to be of interest to users. Various CF schemes have been proposed, but most of them are seriously limited by a cold-start problem which refers to a situation that RSs are incapable of drawing recommendations for new items, new users or both. Moreover,insignificant ratings whose values are less than the corresponding average ratings adversely affect recommendations.In this paper, we propose a Credible and cO-clustering filterBot for cold-stArt recommendations (COBA). It filtersinsignificant ratings by introducing rating confidence level, which substantially reduces the dimensionality of the item-user matrix. To overcome data sparsity, COBA co-clusters items and users, and smoothes ratings within every user cluster. Finally, it predicts user preference byfusing recommendations from item and user clusters. Our experiments show that COBA solves the cold-start problem regarding recommendation accuracy and scalability.