We expose the principles of agglomerative clustering of networks and propose a new efficient link clustering algorithm with a relational constraint, bound implicitly to the corresponding line graph of the input network. Along we develop dissimilarity measures, which besides the network structure consider properties of network elements. We evaluate the algorithm on a set of networks, including bibliographic networks from the field of topological indices. Using existent and new scientometric network analysis approaches we analyze them in detail. We design a method for general hierarchy visualization and develop a visualization method for mobile networks. We use the methods on suitable networks. Considering the principles of abstraction and interactivity we develop a new extendable tool for continuous analysis and visualization of large networks – net.Plexor, which introduces new structured real-time approaches into the network analysis, advanced methods of visualization and upper methods. We conclude the work with an overview of network file formats, and give advice on network data collection and storage.
|
