Hierarchical and massively interactive approaches for hybridization of evolutionary computations and agent systems—comparison in financial application

Abstract

When we think about hybridizing of evolutionary computations and agent systems in fact two approaches are possible: (1) hierarchical one – where agents are used as the management layer and the evolutionary algorithms are executed inside (sub)populations “within” agents and (2) system realized as the population(s) of evolving agents equipped with “DNA” performing life-steps to obtain their life-goals. In this paper we discuss aforementioned approaches and present their sample realization and application for solving a challenging portfolio optimization problem defined as a multi-objective optimization problem with maximization of the investment profit and minimization of the investment risk level.