In most existing recommender systems, implicit or explicit interac- tions are treated as positive links and all unknown interactions are treated as negative links. e goal is to suggest new links that will be perceived as positive by users. However, as signed social net- works and newer content services become common, it is important to distinguish between positive and negative preferences. Even in existing applications, the cost of a negative recommendation could be high when people are looking for new jobs, friends, or places to live.
In this work, we develop novel probabilistic latent factor mod- els to recommend positive links and compare them with existing methods on ve di erent openly available datasets. Our models are able to produce be er ranking lists and are e ective in the task of ranking positive links at the top, with fewer negative links ( ops). Moreover, we nd that modeling signed social networks and user preferences this way has the advantage of increasing the diversity of recommendations. We also investigate the e ect of regularization on the quality of recommendations, a ma er that has not received enough a ention in the literature. We nd that regularization pa- rameter heavily a ects the quality of recommendations in terms of both accuracy and diversity.