Étude comparative des approches de détection de communautés

Abstract

The analysis of complex networks has vast and varied applications, ranging from social networks to biological and technological systems. Community detection is a key task in this analysis, as it allows the division of a network into subgroups of nodes that are densely connected to each other but weakly connected to the rest of the network. This study focuses on unsupervised algorithms for community detection, which are capable of identifying these structures without requiring labeled data. These methods rely on various approaches such as the optimization of modularity, spectral methods, and dynamic processes. The thesis provides a detailed analysis of unsupervised techniques, highlighting their advantages and challenges, particularly regarding scalability and robustness in large networks. The study also presents a comparative evaluation of these algorithms across several types of complex networks to assess their effectiveness and reliability in detecting communities. Finally, potential improvements to existing algorithms are discussed, with a focus on their application in modern, large-scale networks