Improving collaborative filtering with trust-based metrics

Despite its success, similarity-based collaborative filtering suffers from some significant limitations, such as scalability and sparsity. This paper introduces trust to the domain of collaborative filtering to overcome these limitations. Compared with the similarity-based CF, introduction of trust does improve the performance of CF in terms of coverage, prediction accuracy, and robustness in the presence of attacks. Experimental results based on a real dataset are illustrated as evidences to support our claim.