Simuler pour comprendre : un éclairage sur les dynamiques de marchés financiers à l’aide des systèmes multi agents

Bruno BEAUFILS, Olivier BRANDOUY, Lin MA, Philippe MATHIEU


« What are the minimum specifications to design a realistic atificial financial market ?». Starting from this research question, it is argued in this article that simulation of Multi-Agent Systems (MAS) can be a specific mode of inquiry which improves our knowledge of financial motions. Among others, we show that one must pay a particular attention to market architectures implemented in the MAS in casu generating realistic artificial data is targeted. This element, i.e. market microstructure, appears to be necessary and sufficient in explaining the emergence of a wide range of financial stylized facts at the intraday level, while the design of agents behaviors could be considered as of less value for this emergence. This research has potential implications in the design of regulatory financial institutions or algorithmic trading methods. It also makes a point on a new and promising research methodology, at the crossroads of Information Systems, Artificial Intelligence and Finance.


Finance, simulation, microstructure, multi agents systems, artificial intelligence.

Full Text:

 Subscribers Only


Copyright (c)