Abstract
Social networks present a hidden hierarchical organization explaining some of the basic properties of the structure. In fact, the hierarchy structure of networks describes the organization of communities joined to form the entire network. Thus, hierarchical clustering for community discovering present a crucial task. In fact, most of the existing algorithms proposed for hierarchical community detection are either based on agglomerative or divisive process. For the best of our knowledge, there is no approach relying on the combination of these two processes. This paper presents a novel hybrid method based on both optimization and hybrid hierarchical clustering for community structure in social networks. Because hybrid hierarchical model usually generates the problem of convergence to a locally optimal detected community, we proposed a new modularity based opinion leader function and we introduced a metaheuristic namely genetic algorithm to optimize this introduced quality function. In fact, we introduce a genetic hybrid formulation of hierarchical clustering for community detection in social network, where hierarchical bottom-up and top-down methods are combined in order to produce the same community structure. Performances of the proposed approach are evaluated using both real and artificial networks. Experiments show the efficiency of the introduced method in improving the execution time and enhancing the quality of the clustering results.