Density-based Community Identification and Visualisation

Community can be generally defined as a sub graph where nodes are more densely connected with each other than with the rest of a network. Such definition makes application of density-based clustering methods to community identification justified and natural. Moreover, density-based methods have many extensions enabling their application to complex data analysis. Therefore, the analysis of the characteristics of density-based clustering methods in application to community identification is important and valuable. The article presents and evaluates new similarity measures that can be utilised by the approaches to density-based community identification. Several experiments on real life and generated networks are performed to show and explain the differences between these measures and to compare them with other methods. The results show that the new measures improve the quality of analysis and that density-based clustering algorithms can be valuable community identification methods.