Employing computational intelligence techniques and Big Data analytics in a multi-agent system experiment of finance

Abstract

We propose a research within the Agent-Based Computational Finance area. This area has been studying the application of multi-agent system models and computational intelligence techniques in finance. In this context, we intend to employ different computational intelligence techniques (reinforcement learning, genetic algorithm and fuzzy logic) and big data analytics in a multi-agent system experiment (agent-based model simulation of a double auction market) of finance. The research aims are to investigate of the ability of reinforcement learning to model the agents learning and evolution process in the financial markets, to study the use of multi-criteria performance indexes in order to model the agent learning process, and to analyse the consequence of this agents' behaviour to the financial markets as a whole. Our plan is the introduction of different computational intelligence techniques in an agent-based model simulation of the financial market, and the test of a list of hypotheses about the agents and the financial markets micro-structure. This research is relevant because its outcome can help researchers, financial regulators, policymakers and practitioners to better understand the agents learning and evolution process, and its consequences to the financial market micro-structure as a whole. We are requesting fellowships for one postdoctoral researcher with expertise in Computational Finance. In terms of internationalization, this research has an international collaboration with the Institute for Analytics and Data Science and the School of Computer Science and Electronic Engineering of the University of Essex, UK. As a part of this project, we intend to intensify our relationship though a series of researchers' visits and interchange.