Data di Pubblicazione:
2016
Citazione:
Predicting the link strength of "newborn" links / M. Zignani, S.T. Gaito, G.P. Rossi - In: WWW '16 Companion : Proceedings[s.l] : ACM, 2016. - ISBN 9781450341448. - pp. 147-148 (( Intervento presentato al 25. convegno International World Wide Web Conference tenutosi a Montreal nel 2016 [10.1145/2872518.2889367].
Abstract:
Measurements of online social networks (OSNs) support the common fact that not all links carry the same social value, and that the strength of each link is strictly related to the frequency of interactions between the connected users. In this paper, we investigate the predictability of the interactions on OSN links by wondering if it is possible to categorize interactive or non-interactive links at their creation time. We turn the problem into a binary classification task and introduce a set of features which leverage the temporal and topological properties of the social and interaction networks, without requiring the knowledge of the interaction history of the link. The best classifier trained on a Facebook dataset obtained 0.72 as AUC. The above performance suggests that we can distinguish between interactive/non-interactive links at the time of link creation.
Tipologia IRIS:
03 - Contributo in volume
Keywords:
interaction prediction; interaction graphs; online social networks
Elenco autori:
M. Zignani, S. Gaito, G.P. Rossi
Link alla scheda completa:
Titolo del libro:
WWW '16 Companion : Proceedings