Graph Networks as Learnable Engines for Relations Inference of Interacting Financial Systems

Abstract

Although the heterogeneous of financial markets is attracting interest both among scholars and practitioners, however, attention was almost exclusively given to networks in which all individuals were treated indifference, while neglecting all the extra information about the context-related or temporal-spatial properties of the interactions under study. Here introduces a new learnable relation inference model—based on graph networks—which implements an inference for entity- and relation-centric representations of multilayer, dynamical systems. The results show that as a learnable model, the approach supports accurate predictions from real and simulated data.