Cyclic preferential attachment in complex networks

Abstract

Preferential Attachment (PA), which was originally proposed in the Barabasi-Albert (BA) Model, has been widely ac- cepted as a network growth model which returns in scale-free networks. Preferential attachment in the BA model operates on the assumption that a node which has more links has a better likelihood to create new links. In this work, we expand the PA mechanism by treating it as a cyclic mechanism which is linked to both direct and indirect neighbours of a node. The assumption behind this extension is that the preference of nodes is influenced by their indirect neighbours as well. We show that traditional PA can be absorbed as a special case of this new growth model, which we name ‘cyclic preferential attachment’ (CPA). We also discuss the properties of simulated networks that were generated based on CPA. Finally, we compare and contrast the CPA based networks with the traditional PA based networks and several real-world networks of similar sizes and link-to-node ratios, and show that CPA offers more flexibility in modeling real world networks.