Mechanics and Quality of Agent-Informational Clustering in Social Networks

Abstract

The present paper is devoted to the study of the mechanics of agent-informational clustering in a social network on the example of user segmentation tasks taking into account an influence criterion. The main features of data generated by social networks (social big data) and metrics that characterize influential network nodes are considered. A review of community-building algorithms based on the theory of social networks, as well as clustering methods based on machine learning, is carried out. Metrics for assessing the quality of segmentation are presented. The results of the application of methods (selected on the basis of the performed analysis) to a test dataset are shown. The limitations of the applicability of considered approaches and possible problems during the implementation of algorithms in the field of social network analysis are described. Evaluation of the effectiveness is performed.