Recommending collaboration links in coauthorship networks

Žontar, Luka (author), Curk, Tomaž (mentor)

In the evolving landscape of academic research, interdisciplinary and cross-institutional collaboration has become critical. However, identifying valuable partnerships remains a complex task. This thesis introduces a recommender system for suggesting new academic collaborations within coauthorship networks, focusing on promoting cross-institutional research within the EUTOPIA alliance. We construct a large-scale coauthorship dataset from bibliographic sources such as Elsevier, Crossref, and ORCID, and model it using graph neural networks. Our work involves an ablation study of various GNN backbones, including LightGCN, GraphSAGE, and attention-based models, combined with different loss functions and node feature sets. The final model uses weighted temporal embeddings of article content and achieves MRR@10=19.0% ± 0.3% and HitRate@10=36.4% ± 0.5%, outperforming the baseline methods. Although attempts to incorporate keyword popularity as a feature were inconclusive, our work highlights the potential of GNN-based recommender systems to enhance research collaboration networks. In addition, we provide a detailed analysis of collaboration dynamics, highlighting the importance of new research collaborations and investigating how research trends, research interests, and lead authors affect the occurrence of new collaborations.

• Language: English
• Keywords: graph neural networks, link prediction, homogeneous networks, bibliography mining, scientometrics
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2025
• PID: 20.500.12556/RUL-176013
• COBISS.SI-ID: 259879683
• Publication date in RUL: 18.11.2025

Secondary language

• Language: Slovenian
• Title: Priporočanje sodelovalnih povezav v soavtorskih omrežjih

Abstract

V sodobnem akademskem raziskovalnem okolju so interdisciplinarne in medinstitucionalne raziskave postale ključnega pomena, vendar je iskanje primernih raziskovalnih partnerstev pogosto kompleksno. V magistrskem delu predstavimo priporočilni sistem za predlaganje novih sodelovanj v okviru soavtorskih omrežij, s poudarkom na spodbujanju medinstitucionalnega sodelovanja znotraj evropske univerzitetne zveze EUTOPIA. Najprej zgradimo obsežno podatkovno zbirko iz bibliografskih virov, kot so Elsevier, Crossref in ORCID, ter jo modeliramo z grafnimi nevronskimi mrežami. Naše delo vključuje ablacijsko študijo različnih modelov GNN, vključno z arhitekturami LightGCN, GraphSAGE, GATv2 in Transformer, kombiniranimi z različnimi funkcijami izgube in tehnikami vložitev. Končni model uporablja vložitve akademskih člankov utežene glede na nedavnost objave članka in doseže MRR@10=19,0% ± 0,3% ter HitRate@10=36,4% ± 0,5%, s čimer preseže osnovne pristope. Kljub temu da vključitev indeksa priljubljenosti ključnih besed kot značilke, ki ga vpeljemo v tem delu, ni pokazala izboljšav, naše delo izpostavlja potencial grafnih nevronskih mrež kot priporočilni sistem za krepitev raziskovalnega sodelovanja. Poleg tega v delu podamo podrobno analizo dinamike raziskovalnega sodelovanja, pri čemer izpostavimo pomen novih raziskovalnih sodelovanj ter preučimo, kako raziskovalni trendi, raziskovalni interesi in vodilni avtorji vplivajo na nastanek novih sodelovanj.

• Keywords: grafne nevronske mreže, napovedovanje povezav, homogena omrežja, razvijajoče se mreže, rudarjenje bibliografije, scientometrija