Uma Abordagem para Detecção de Comunidades a partir de Sequências de Interações Sociais

Abstract

The topology of a social network and the temporal aspect of the interactions between a pair of nodes indicate the strength of the relationship between them and allow to classify it. For example, a relationship can be classified as persistent and embedded based, respectively, on the regularity with which interactions occur and on the number of neighbors in common among the pair of nodes involved. On the other hand, a rare and little embedded relationship is random and represents noise in a social network, hiding the most significant structure of the network and preventing an accurate analysis. In this work, we propose a framework to handle social network data that exploits temporal and topological features of its sequences of real and synthetic interactions to improve the detection of static communities by existing algorithms. By removing random relationships, we observe through multiple sources of evidence that social networks converge to a topology with more purely social relationships and higher quality community structures.