Clustering and cycles in bipartite networks: an application to social systems.
By: Pedro Lind
From: Institute for Computational Physics,
Universität Stuttgart, Germany
At: Complexo Interdisciplinar, Anfiteatro
[2006-06-21]
We investigate the clustering coefficient in bipartite networks where cycles of size three are absent and therefore the standard definition of clustering coefficient cannot be used. Instead, we introduce a simple coefficient given by the fraction of cycles with size four and show that both coefficients yield the same clustering properties. The new coefficient is computed for two networks of sexual contacts, one bipartite and another where no distinction between the nodes is made (monopartite). Combining both clustering coefficients we deduce an expression for estimating cycles of larger size, which improves previous estimations and is suitable for either monopartite and multipartite networks. In addition, discussing the applicability of such analytical estimations, a recent model for social networks based in systems of colliding particles is described.