Garrido Yuste, Guillermo2024-05-202024-05-202010-09-01https://hdl.handle.net/20.500.14468/14682In this work, we study the community structure of endorsement networks, i.e., social networks in which a directed edge u → v is asserting an action of support from user u to user v. Examples include scenarios where a user u is favouring a photo, liking a post, or following the microblog of user v. Very often, endorsement networks are sub- networks of more complex social systems; for instance, a photo-sharing site typically includes a “favouring” function, which induces an endorsement network. We start from the hypothesis that the footprint of a community in an endorsement network is a bipartite directed clique from a set of followers to a set of leaders, and apply frequent itemset mining techniques to discover such bicliques. Our analysis of real networks indicated that, with high statistical significance, this hypothesis holds, and that the leaders of a community are endorsing each other forming a very dense nucleus. Our method produces many similar bicliques, which are different footprints of the same community. Thus, we propose a novel clustering technique in order to coalesce similar bicliques into meaningful communities. We explore different similarity measures based on set similarity and on edge density between followers and leaders, and by expressing edge density as an inner product operation we show how to make the clustering algo- rithm scalable. Our experiments demonstrate that our clustering algorithm is capable of discovering communities characterised by a set of leaders who link to each other and followers who link to the leaders.eninfo:eu-repo/semantics/openAccesstesis de maestría