Incorporating Tie Strength in Robust Social Recommendation

In this paper, we present a novel method in making recommendations by leveraging Tie Strength, an integrated social relationship measurement calculated from various user information gathered from social media. Moreover, the proposed method adopts Least Absolute Errors in factorization scheme to reduce the sensitivity to data outliers. We have conducted comprehensive experiments over the real datasets from popular social media services. The evaluation results demonstrate that the proposed method outperforms certain state-of-the-art social recommendation methods in terms of Root Mean Squared Error and Precision versus Recall measures.