On the Design of Agent-Based Artificial Stock Markets - Agents and Artificial Intelligence - page 345

On the Design of Agent-Based Artificial Stock Markets

Olivier Brandouy 1 , Philippe Mathieu 2 , and Iryna Veryzhenko 3

1 Sorbonne Graduate Business School, Paris, France
2 LIFL, UMR CNRS-USTL 8022, Lille, France
3 LEM, UMR CNRS 8174, Lille, France
[e-mail],
[e-mail],
[e-mail]

Abstract. The purpose of this paper is to define software engineering abstrac-
tions that provide a generic framework for stock market simulations. We demon-
strate a series of key points and principles that has governed the development
of an Agent-Based financial market application programming interface (API).
The simulator architecture is presented. During artificial market construction we
have faced the whole variety of agent-based modelling issues : local interaction,
distributed knowledge and resources, heterogeneous environments, agents auton-
omy, artificial intelligence, speech acts, discrete or continuous scheduling and
simulation. Our study demonstrates that the choices made for agent-based mod-
elling in this context deeply impact the resulting market dynamics and proposes a
series of advances regarding the main limits the existing platforms actually meet.

Keywords. Multi-agent systems, Artificial market, Market microstructure,

Agents behaviour.

1 Introduction

Multi-agent modelling is nowadays actively applied to financial markets simulation.
This is partly due to its ability at reflecting a wide range of complexities arising in these
markets, and to its flexibility for exploring the impact of automation in trading. Thus,
agent-based computational simulations [1] may contribute to several scientific debates
in Finance. One example at the crossroads of multi-agent modelling and machine learn-
ing is Bayesian learning : this technique can be used by agents to incorporate all avail-
able information into the decision making process [2]. It can also be employed to track
a moving parameter [3] such as the fundamental value of a given stock. On the other
hand, financial markets offer an important field of application for agent-based modelling
and machine learning, since agent objectives and interactions are clearly defined. For
this reason, financial market environments can help to answer some modelling issues
related to agent engineering, or test robustness of existing behavioural models.
At the present, there is a large number of agent-based frameworks, with varying func-
tionalities and architectures, addressing different problems. Generally speaking, there
are two major approaches to agent-based financial market simulations. The first one is
leads to the realization of a specific market structure, with specific agents' behaviours,