Finding suitable number of recommenders for trust-aware recommender systems: An experimental study

Abstract

Finding suitable number of recommenders helps to improve the efficiency and the rating prediction accuracy in the trust-aware recommender system (TARS). Most existing works involve all available recommenders to enlarge the diversity of recommendations. However, it is computational expensive if the number of involved recommenders is big. In this work, we experimental study real application data to find the suitable number of recommenders needs to be involved in TARS. It is suggested that all recommenders should be involved for most users. For those who have sufficient number of recommenders, only part of recommenders is suggested to be involved, i.e., around one fourth of the maximum number of recommenders in our experimental dataset.