M. Pelegrin, E. Carrizosa, A. Marín
In recent years, community discovery led by key nodes identification has become a hot research topic in network analysis. Since real networks usually display a modular structure, it is commonly assumed that targeted key members will be in different compartments or communities. Recent works identify key nodes in a first step and use them as communities centroids afterwards. However, none of them consider joint optimization of key nodes and communities. This work presents a mathematical programming formulation for identifying the group of most relevant nodes and their spheres of infuence, interpreted as communities. The network clustering
yielding maximum overall relevance of the clusters centroids, calculated as an eigenvector centrality, is selected. The proposed exact approach serves as a suitable adaptation of widespread PageRank to the problem of group centrality. Our computational tests show its potential to uncover complex network structures, in a context where heuristics abound.
Palabras clave: Network Analysis, Eigenvector centrality, Clustering, Mathematical Optimization
Programado
RM-2 Premio Ramiro Melendreras
3 de septiembre de 2019 18:30
I3L1. Edificio Georgina Blanes