Exploiting Unlabeled Ties for Link Prediction in Incomplete Signed Networks

摘要

Signed networks analysis has attracted increasing attention in the areas of friendship recommendation and trust relationship prediction. Existing methods for link prediction in signed networks tend to rely on the features only from labeled ties which are not ideal for handling incomplete networks. In order to solve this issue, in this paper, we present a novel model called UTLP (Unlabeled Ties based Link Prediction) for incomplete signed networks. Based on social awareness theories, UTLP can utilize the features of both labeled ties and unlabeled ties. We consider four types of features including node-based features, structural balance-based features, status-based features and latent features, which can provide abundant evidence to improve the accuracy of prediction. With the extracted features, we adopt transfer learning algorithm with instance weighting to utilize more useful training instances in the source network to train the model and predict the signs of links, which can effectively make up for the incompleteness of target samples. We conduct experimental studies based upon Epinions, Slashdot and Wikipeida. Experiments demonstrate the effectiveness and the efficiency of our proposed approach.