Yousaf, Farhad (2024) Clustering of YouTube Channels by means of Optimal Transport Methods. [Tesi di dottorato]

Documento PDF Thesis_Farhad_Yousaf.pdf Visibile a [TBR] Amministratori dell'archivio Download (3MB) | Richiedi una copia

Abstract

In this manuscript, we focus on the analysis and comparison of complex networks, from the point of view of the optimal transportation theory. We specifically examine the potential applications of the Fused Gromov-Wasserstein distance, a cutting-edge measure that accounts for both feature-based and structural similarities between networks simultaneously. Using the FGW distance, we can consider not only the topological connections across networks, but also node information like content kind and subscriber count. The primary goals of this work are to apply clustering techniques to mining patterns from network data and to examine similarities among random network realizations. The project’s overarching goals include applying clustering algorithms to identify patterns in network data and evaluating similarities for various random network realizations. We will focus on partitioning YouTube networks into meaningful clusters and on the identification of structural patterns of shared material and viewer engagement. To this aim we utilize spectral clustering on the FGW distance matrix. The results show that FGW distance is a very effective tool for network discrimination and has the potential to be a powerful analytical tool for complex networks across many domains. In short, this study sheds light on how optimal transport theory applies clustering techniques to uncover hidden structures and relationships from large amounts of network data, and how it more accurately uses the Fused Gromov-Wasserstein distance to assess network similarity. This establishes the framework for future investigations that will compare various networks, from biological and communication systems to social media applications

Downloads

Actions (login required)

Modifica documento