Abstract: The ongoing exponential growth of the Internet brings an information overload, which greatly increases the necessity of effective recommender systems for information filtering. However, collaborative filtering, which is recognized as the most successful technique in designing recommender systems, still encounters the data sparsity problem. Social relations have been found to be effective to improve the prediction accuracy of recommender systems. In order to handle the data sparsity problem, this paper proposed a new social matrix factorization recommender algorithm by leveraging the Logistic function. Experimental results on three real-world datasets illustrate that the proposed method provides more accurate recommendation results, especially under sparse conditions.