Recommending Positive Links in Signed Social Networks by Optimizing a Generalized AUC

Dongjin Song, David A. Meyer

Abstract: With the rapid development of signed social networks in which the relationships between two nodes can be either positive (indicating relations such as like) or negative (indicating relations such as dislike), producing a personalized ranking list with positive links on the top and negative links at the bottom is becoming an increasingly important task. To accomplish it, we propose a generalized AUC (GAUC) to quantify the ranking performance of potential links (including positive, negative, and unknown status links) in partially observed signed social networks. In addition, we develop a novel link recommendation algorithm by directly optimizing the GAUC loss. We conduct experimental studies based upon Wikipedia, MovieLens, and Slashdot; our results demonstrate the effectiveness and the efficiency of the proposed approach.