Predicting Positive and Negative Links in Signed Social Net- works via Transfer Learning

Different from a large body of research on social networks that almost exclusively focused on positive relationships, we study signed social networks with both positive and negative links. Specifically, we focus on how to reliably and effectively predict the signs of links in a signed social network (called a target network), where a very small amount of edge sign information is available as the training data. To train a good classifier, we adopt the transfer learning approach to leverage the abundant edge signs from another signed social network (called a source network) which may have a different joint distribution of the observed instance and the class label. As there is no predefined feature vector for the edge instances ii in a signed network, we construct generalizable features that can transfer the topological knowledge from the source network to the target. With the extracted features, we adopt an AdaBoost-like transfer learning algorithm with instance weighting to utilize more useful training instances in the source network for model learning. Experimental results on two real large signed social networks demonstrate that our transfer learning algorithm can improve the prediction accuracy by 40 % over baseline schemes. ii