Fouilles de graphes et Maximisation de l’Influence dans les réseaux sociaux

Capture d’écran 2026-07-14 à 01.01.50

Soutenue par Motaz BENHASSINE, en co-tutelle entre l’Université d’Artois et l’Université de Monastir (Tunisie) le 13-12-2024. Co-Encadrée par Said Jabbour, Mohamed Graiet. Cadre magnifique avec un bel accueil. Très bonne thèse à lire. [theses.fr] [rapporteur & président du jury]

In this thesis, we focus on the development of clustering methods and algorithms for detecting disjoint communities and using them for influence maximization. More specifically, we propose new similarity measures and study their impact on community detection in terms of complexity. The second part of this thesis involves analyzing the impact of community quality on influence maximization in social networks. Starting from the detected communities, our goal is to determine to what extent the quality of these communities affects information propagation under the same seed nodes selection criteria and the same diffusion model. The final part of the thesis involves studying hybrid approaches to solving the conceptual clustering problem and their impact on community detection. More precisely, we define new patterns called disjunctive itemsets that can help achieve more relevant clusters. SAT encodings are proposed for enumerating these patterns using propositional logic. Finally, experiments on a wide range of graphs and various transactional datasets were conducted to validate our approaches and demonstrate the effectiveness of our models.

Laisser un commentaire