Algorithm for searching for sporadic contextual community in internet social networks

Abstract

The presented paper describes an approach to finding sporadic contextual communities in Internet social networks. The developed algorithm and software combine an attribute-based method for agent identification, context-based message analysis, and a graph-based method for evaluating their interactions. This approach will automate the selection of target user groups on the Internet for analytical purposes, dissemination of targeted information, and identification of influence in communities. The paper highlights the importance of sporadic contextual communities in social networks that share common characteristics and behaviors but are not explicitly unified. The proposed approach relies on analyzing graph structure and vertex attributes, taking into account message context, dynamics, and other constraints. Using the hybrid approach in analyzing sporadic communities will effectively identify target user groups, enhance analytical research capabilities, and identify constructive and destructive influences in communities. The developed method is an innovative tool for automating the process of analyzing social networks and identifying key communities for further research and action. Analysis of the results showed that all agents found did not belong to known communities, the accuracy of SCS identification was about 70%. One-time meetings were noted, probably due to the short-term nature of the observation, but the hypothesis about the presence of influential participants with a large number of connections was confirmed. The experimental study identified key features of user behavior in the social network and confirmed the hypothesis of sporadic contextual communities. The results of the study can be used to improve social network analysis algorithms and increase the accuracy of key communities’ identification.