Using Preference Intensity for Detecting Network Communities

Abstract

In a real-world network, overlapping structures are essential for understanding the community. In many different situations, a node may join or leave, and this defines sub-communities of varying size. In this paper, we propose a preference implication based-method for generating overlapping structures based on a local function optimization approach. We introduce some parameters in our novel method to design the communities according to a threshold. This method allows us to control the size and number of these overlapping regions. The $\nu$ will enable us to design the sub-communities. This framework can easily detect communities in a scale-free network case. We set our experiments using artificial and real network data with a size between ${\approx }15$ to ${\approx }10000$. In our findings, we found a good relationship between $\nu$ and overlapping nodes in communities. We control our procedure using $\alpha$ parameter as well. We can say that the preference is stronger when $\nu$ is greater than 0.5, and a value of $\alpha$ between 0.20 and 0.80. The third parameter $\delta$, which controls the intensity of community membership, defines the degree of relationship of a node to a community. The communities detected by the preference implication method obey a power law in the community size distribution.