A probabilistic model for networks generated by actors’ characteristics

Abstract

This work presents the affinity network model for random graphs, consisting of a broad family of random graph models depending on some parameters. In this model, we suppose that each individual randomly chooses a set of characteristics that represent him according to a certain probability measure. The connections between two individuals depend on their shared characteristics and are valued according to a function that measures what we call affinity in the network. According to the choice of this function, the network’s density can vary from sparse to complete graph, causing the model to be very flexible, which makes it suitable to fit with real networks. To illustrate the behavior of the affinity network model, we present a Monte Carlo simulation study. We tune the model’s generating parameters, analyze its topological measurements, and compare them with equivalent graphs with edges occurring independently, disregarding actors’ characteristics.