The Identification for Participants of Computer Networks by Modified Clustering Method

摘要

The analysis of the network participants is carried out on the basis of graph models and allows defining the associated groups. For find groups in networks, the clustering methods in graph models are widely used: method of centrality, k-means, modularity, hierarchical, spectral, and partitioning around medoids. In such methods of clustering, the results essentially depend on the selected metric. We proposed modifications to the clustering method based on the generalized centrality metric, which takes into account many properties of network participants. In the proposed generalized metric, we applied properties for the centrality. For analyze the accuracy of the proposed method, we considered a group of network participants, which consists of three hundred participants. Comparison of the results of clustering by the proposed method on the basis of a generalized metric shows the accuracy in the definition of groups of network participants similar to those of methods: spectral, k-means, modularity and superior in accuracy methods: centrality, hierarchical.