A Theory-Driven Approach for Link and Unlink Predictions in Directed Social Networks[go to overview]
Predicting and explaining the formation and dissolution of ties in social networks requires a deep understanding of the sociological factors that drive the linkage dynamics in social systems. In this paper, we present a computational social science approach that translates sociological concepts to structural measures on directed link and unlink networks. We demonstrate the utility of our approach in an empirical case study on the social network of German politicians from before, during and after the German federal election. Finally, we show which measures are suited to predict the formation and dissolution of social ties in this network and quantify the added value of unlinking information for both prediction tasks. Our results show amongst others that the social tie formation behavior of a user is more in line with the tie formation behavior of his friends or group members than his tie dissolution behavior. Further, our results suggest that measures based on information about past links are extremely valuable for predicting the dissolution of social ties, while for the prediction of the formation of social ties measures based on the link network are sufficient.
25.09.14 - 08:15