A threshold method for financial networks and geometric scattering of agents

Abstract

Various filtering methods are applied to financial networks to overcome dimensionality problem of financial networks. In this study, we present a new threshold method based on Triangular Maximally Filtered Graph (TMFG) filtration. The core idea is to use TMFG to extract a filtration that preserves discrete geometry of the underlying space from the correlation distance matrix, and then combine it with the threshold network to generate a graph model of a financial network without isolated vertices or subgraph. Moreover, we present two new network measures to analyze structural changes of a financial network throughout an economic crisis. © 2019, © 2019 Taylor & Francis Group, LLC.