Information flux in complex networks: Path to stylized facts

Abstract

A market model where agents behave according to endogenous interests and limited attention is proposed. The information diffusion process is based on the model developed from empirical observation of Twitter microblogging platform (Weng et al., 2012). Four types of contact network models are utilized: Albert–Barabási, Erdös–Rényi, circular regular, and a power-law network model proposed here based on Zipf distribution. External influences are modeled as Gaussian distributed inputs. Heavy-tailed return distributions are found for Albert–Barabási and a class of Zipf networks. Multifractal detrended fluctuation analysis indicates that nonlinear correlations in price series are stronger for the same networks. Such nontrivial statistics are also present in information flux proxies. The ability of the model to provide stylized facts usually found in financial data from Gaussian excitation depends on the network topology.