Biclustering neighborhood-based collaborative filtering method for top-n recommender systems

We propose a novel collaborative filtering method for top-\(n\) recommendation task using bicustering neighborhood approach. Our method takes advantage of local biclustering structure for a more precise and localized collaborative filtering. Using several important properties from the field of Formal Concept Analysis, we build user-specific biclusters that are “more personalized” to the users of interest. We create an innovative rank scoring of candidate items that combines local similarity of biclusters with global similarity. Our method is parameter-free, thus removing the need for tuning parameters. It is easily scalable and can efficiently make recommendations. We demonstrate the performance of our algorithm using several standard benchmark datasets and two paypal (in-house) datasets. Our experiments show that our method generates better recommendations compared to several state-of-the-art algorithms, especially in the presence of sparse data. Furthermore, we also demonstrated the robustness of our approach to increasing data sparsity and the number of users.