Influence spreading model in analysing ego-centric social networks

Abstract

In an earlier study one of us had developed a model of influence spreading for analysing human behaviour and interaction with others in a social network. Here we apply this model and corresponding influence centrality measures to real data of mobile phone call detail records. From this we get structures of human ego-centric networks and use a simple model, based on the number of phone calls, to describe the strengths of social relationships. To analyse 48,000 egos in their ego-centric networks we define normalised out-centrality and in-centrality influence measures, by dividing with out-degree and in-degree, respectively. With these and the betweenness centrality measures, we analyse the influence spreading in the ego-centric networks under different scenarios of link strengths between individuals reflecting the network structure being either interaction or connectivity oriented. The model reveals characteristics of social behaviour that are not obvious from the data analysis of raw empirical data or from the results of standard centrality measures. A transition is discovered in behaviour from young to older age groups for both genders and in both normalised out-centrality and in-centrality as well as betweenness centrality results.