Social Networks Modelling
Gascón Marzo, Andreu
Sikora, Anna tut. (Universitat Autònoma de Barcelona. Departament d'Arquitectura de Computadors i Sistemes Operatius)

Additional title:	Modelatge de xarxes socials
Additional title:	Modelaje de Redes Sociales
Date:	2026
Abstract:	Predicting user interactions in decentralized social networks requires capturing both complex community structures and rapidly evolving temporal dynamics. This study presents a scalable Spatio-Temporal Graph Attention framework for multi-task link prediction, utilizing a massive dataset from the Bluesky social protocol. We introduce a Snowball-Tiered Sampling methodology that condenses 11 billion records into a representative million-scale environment while preserving the network’s Power Law characteristics. Our architecture integrates 384-dimensional MiniLM semantic embeddings directly into a heterogeneous graph flow, allowing the model to prioritize topical alignment over simple proximity. By employing a dual-attention mechanism—spatial GAT layers for neighborhood context and temporal self-attention for behavioral "velocity"—the model achieves high predictive precision for the next day interactions.
Rights:	Aquest document està subjecte a una llicència d'ús Creative Commons. Es permet la reproducció total o parcial, la distribució, i la comunicació pública de l'obra, sempre que no sigui amb finalitats comercials, i sempre que es reconegui l'autoria de l'obra original. No es permet la creació d'obres derivades.
Language:	Anglès
Document:	Treball final de grau
Subject:	Graph Neural Networks ; Social Network Modelling ; Attention Network ; Bluesky

16 p, 2.2 MB

The record appears in these collections:
Research literature > Bachelor's degree final project