A generalized model for social collaboration networks

Abstract

In this paper, we constructed a three-parameter general social collaboration network model (GCNM) using both similarity-based strategy and copying-based strategy. In this model, a connection can be introduced between two nodes if they are similar in attributes or similar in structure (i.e., have many common friends). This model can match various types of real-world social networks by just choosing different values for three parameters, and therefore, can reproduce observed real network characteristics. Particularly, three widely studied models (similarity-based, copying-based, and pure preferential attachment (PA)) are only the special cases of this new model. Specifically, this new model possesses following new characteristics which existing models have difficulty to fulfill in all: 1) it not only exhibits the high average clustering coefficient, but also the high global clustering coefficient and explicit community structure (great number of triangles); 2) it not only makes the connections between the new node and existing nodes, but also between existing nodes; and 3) it has not only linear, but also non-linear relationship between the linking probability and the degree (the connectivity of nodes). Numerical results, which are in good agreement with real-world datasets from different fields, demonstrated all of these.