Trust-aware recommender systems are programs that use the friendship connections of users in social networks to provide personalized recommendations. Most of the existing recommendation methods suffer from coldstart and sparsity issues. To overcome these issues in this paper a novel trust-aware recommendation method is proposed. The proposed method groups similar users based on their influential values in social networks which are computed by applying a random walk-based method. Then those of top high influential users are identified as cluster seeds. Each cluster seed is associate with a unique label and then a novel label propagation method is used to propagate the labels to unassigned users. Identified clusters are finally employed in the prediction process to predict missing rating values. Several experiments were performed on two real world datasets to evaluate the performance of the proposed method. The results show the superiority of the proposed method in comparison with traditional and state-of-the-art recommenders
