Résumé:
Community detection in networks plays an essential role in understanding their structures.
The application of machine learning methods to community detection tasks in complex
networks has attracted sustained attention in recent years, we propose in this thesis a new
community detection approach based on k-means with the initialization of central nodes
according to their densities and their degrees, the choice of the number of community k
is made according to the best modularity.
Our approach is efficient, simple and easy to implement. We compared our algorithm
with some state-of-the-art algorithms on synthetic networks and real networks, with the
evaluation measure : Modularity Q, and we obtain very acceptable results.
