Investigating microstructure patterns of enterprise network in perspective of ego network

Abstract

In social networks the behavior of individuals can be researched through the evolution of the microstructure. As we know, triad is the basic atom shape to build the whole social network. However we find that quad plays the basic role rather than triad in Enterprise Network (EN). In particular, we focus on four typical microstructure patterns including triad, 4-cycle, 4-chordalcycle and 4-clique in EN. We propose algorithms to mine these microstructure patterns and compute the frequencies of each type of microstructure patterns in an efficient parallel way. We also analyze the structural features of these microstructure patterns in a perspective of ego network. Additionally we present the evolutionary rules between these microstructure patterns based on the statistical analysis. Finally we combine the features into traditional methods to solve the link prediction problem. The results show that these features and our combination methods are effective to predict links between enterprises in EN.