New approach for graph based data mining model

Abstract

This paper presents new approach of Communication Network Analysis (CNA) that is interdisciplinary subfield of advanced concept of important Social Network Analysis (SNA). Objects in CNA are members of network discovered as vertices that are linked by edges. Identification of relevant vertices within connected components in telecommunication network graphs, such as influencers are proposed. Beside this result, the algorithm describes behaviour between component members, research interactions between components and telecom services usage. Algorithm is based on a combination of two important machine learning techniques - Classification technique Extreme gradient boosting (XGB) and Graph algorithm that consists from Pruning, K-Neighbourhood, Isolated islands and Centrality measure calculation. This data mining model is used in telecommunication companies as part of marketing strategies and campaign management processes since influencers are awarded for contribution in network services spreading and adopting between members.