Forming time-stable homogeneous groups into Online Social Networks

In this work we investigate on the time-stability of the homogeneity - in terms of mutual users' similarity within groups - into real Online Social Networks by taking into account users' behavioral information as personal interests. To this purpose, we introduce a conceptual framework to represents the time evolution of the group formation in an OSN. The framework includes a specific experimental approach that has been adopted along with a flexible, distributed algorithm (U2G) designed to drive group formation by weighting two different measures, mutual trust relationships and similarity, denoted by compactness. An experimental campaign has been carried out on datasets extracted from two social networks, CIAO and EPINIONS, and the results show that the time-stability of similarity measure for groups formed by the algorithm U2G based on the sole similarity criterion is lower than that of groups formed by considering similarity and trust together, even when the weight assigned to the trust component is small.