Link Prediction in Time Varying Social Networks
Predicting new links in complex networks can have a large societal impact. In fact, many complex systems can be modeled through networks, and the meaning of the links depend on the system itself. For instance, in social networks, where the nodes are users, links represent relationships (such as acqu...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-03-01
|
| Series: | Information |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2078-2489/13/3/123 |
| _version_ | 1797470660524507136 |
|---|---|
| author | Vincenza Carchiolo Christian Cavallo Marco Grassia Michele Malgeri Giuseppe Mangioni |
| author_facet | Vincenza Carchiolo Christian Cavallo Marco Grassia Michele Malgeri Giuseppe Mangioni |
| author_sort | Vincenza Carchiolo |
| collection | DOAJ |
| description | Predicting new links in complex networks can have a large societal impact. In fact, many complex systems can be modeled through networks, and the meaning of the links depend on the system itself. For instance, in social networks, where the nodes are users, links represent relationships (such as acquaintance, friendship, etc.), whereas in information spreading networks, nodes are users and content and links represent interactions, diffusion, etc. However, while many approaches involve machine learning-based algorithms, just the most recent ones account for the topology of the network, e.g., geometric deep learning techniques to learn on graphs, and most of them do not account for the temporal dynamics in the network but train on snapshots of the system at a given time. In this paper, we aim to explore Temporal Graph Networks (TGN), a Graph Representation Learning-based approach that natively supports dynamic graphs and assigns to each event (link) a timestamp. In particular, we investigate how the TGN behaves when trained under different temporal granularity or with various event aggregation techniques when learning the inductive and transductive link prediction problem on real social networks such as Twitter, Wikipedia, Yelp, and Reddit. We find that initial setup affects the temporal granularity of the data, but the impact depends on the specific social network. For instance, we note that the train batch size has a strong impact on Twitter, Wikipedia, and Yelp, while it does not matter on Reddit. |
| first_indexed | 2024-03-09T19:40:11Z |
| format | Article |
| id | doaj.art-1af0f4b1b23b4c6787a10d46f0da455f |
| institution | Directory Open Access Journal |
| issn | 2078-2489 |
| language | English |
| last_indexed | 2024-03-09T19:40:11Z |
| publishDate | 2022-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Information |
| spelling | doaj.art-1af0f4b1b23b4c6787a10d46f0da455f2023-11-24T01:41:35ZengMDPI AGInformation2078-24892022-03-0113312310.3390/info13030123Link Prediction in Time Varying Social NetworksVincenza Carchiolo0Christian Cavallo1Marco Grassia2Michele Malgeri3Giuseppe Mangioni4DIEEI-Dipartimento di Ingegneria Elettrica, Elettronica Informatica, Universitá degli Studi di Catania, I95125 Catania, ItalyDIEEI-Dipartimento di Ingegneria Elettrica, Elettronica Informatica, Universitá degli Studi di Catania, I95125 Catania, ItalyDIEEI-Dipartimento di Ingegneria Elettrica, Elettronica Informatica, Universitá degli Studi di Catania, I95125 Catania, ItalyDIEEI-Dipartimento di Ingegneria Elettrica, Elettronica Informatica, Universitá degli Studi di Catania, I95125 Catania, ItalyDIEEI-Dipartimento di Ingegneria Elettrica, Elettronica Informatica, Universitá degli Studi di Catania, I95125 Catania, ItalyPredicting new links in complex networks can have a large societal impact. In fact, many complex systems can be modeled through networks, and the meaning of the links depend on the system itself. For instance, in social networks, where the nodes are users, links represent relationships (such as acquaintance, friendship, etc.), whereas in information spreading networks, nodes are users and content and links represent interactions, diffusion, etc. However, while many approaches involve machine learning-based algorithms, just the most recent ones account for the topology of the network, e.g., geometric deep learning techniques to learn on graphs, and most of them do not account for the temporal dynamics in the network but train on snapshots of the system at a given time. In this paper, we aim to explore Temporal Graph Networks (TGN), a Graph Representation Learning-based approach that natively supports dynamic graphs and assigns to each event (link) a timestamp. In particular, we investigate how the TGN behaves when trained under different temporal granularity or with various event aggregation techniques when learning the inductive and transductive link prediction problem on real social networks such as Twitter, Wikipedia, Yelp, and Reddit. We find that initial setup affects the temporal granularity of the data, but the impact depends on the specific social network. For instance, we note that the train batch size has a strong impact on Twitter, Wikipedia, and Yelp, while it does not matter on Reddit.https://www.mdpi.com/2078-2489/13/3/123deep learninggeometric deep learninginformation spreadinglink predictionsocial networks |
| spellingShingle | Vincenza Carchiolo Christian Cavallo Marco Grassia Michele Malgeri Giuseppe Mangioni Link Prediction in Time Varying Social Networks Information deep learning geometric deep learning information spreading link prediction social networks |
| title | Link Prediction in Time Varying Social Networks |
| title_full | Link Prediction in Time Varying Social Networks |
| title_fullStr | Link Prediction in Time Varying Social Networks |
| title_full_unstemmed | Link Prediction in Time Varying Social Networks |
| title_short | Link Prediction in Time Varying Social Networks |
| title_sort | link prediction in time varying social networks |
| topic | deep learning geometric deep learning information spreading link prediction social networks |
| url | https://www.mdpi.com/2078-2489/13/3/123 |
| work_keys_str_mv | AT vincenzacarchiolo linkpredictionintimevaryingsocialnetworks AT christiancavallo linkpredictionintimevaryingsocialnetworks AT marcograssia linkpredictionintimevaryingsocialnetworks AT michelemalgeri linkpredictionintimevaryingsocialnetworks AT giuseppemangioni linkpredictionintimevaryingsocialnetworks |