In view of the exponential growth of information, a personalized recommendation has been a critical approach to solving the information overload problem recently. As one of the widest applied recommendation methods, Regularized Singular Value Decomposition (RSVD) conveniently fits the user-item rating matrix by low-rank approximation from explicit user feedback. However, implicit information is also very effective in improving recommendation algorithms, such as the degree correlation of the user-item bipartite network. Consequently, in this paper, we propose a hybrid collaborative filtering model named RSVD_WNBI. It builds on the algorithm RSVD which involves the explicit influence of ratings, and further integrates implicit influence of the degree correlation in the user-item bipartite network from Weighted Network-Based Inference (WNBI) algorithm. Experimental results on three real-world datasets show that our algorithm can yield better performance over already widely used methods in the accuracy of recommendation, especially when few user ratings are observed.