Lucas Iris, Cotsaftis Michel, Bertelle Cyrille
Normandie Univ, UNIHAVRE, LITIS, Le Havre, 25 Rue Philippe Lebon, 76600 Le Havre, France.
ECE Paris School of Engineering, 37 Quai de Grenelle, 75015 Paris, France.
Chaos. 2018 Feb;28(2):023114. doi: 10.1063/1.5010185.
This paper introduces an original mathematical description for describing agents' decision-making process in the case of problems affected by both individual and collective behaviors in systems characterized by nonlinear, path dependent, and self-organizing interactions. An application to artificial financial markets is proposed by designing a multi-agent system based on the proposed formalization. In this application, agents' decision-making process is based on fuzzy logic rules and the price dynamics is purely deterministic according to the basic matching rules of a central order book. Finally, while putting most parameters under evolutionary control, the computational agent-based system is able to replicate several stylized facts of financial time series (distributions of stock returns showing a heavy tail with positive excess kurtosis, absence of autocorrelations in stock returns, and volatility clustering phenomenon).
本文介绍了一种原创的数学描述，用于描述在具有非线性、路径依赖和自组织相互作用的系统中，受个体和集体行为影响的问题中主体的决策过程。通过基于所提出的形式化设计一个多主体系统，提出了在人工金融市场中的应用。在这个应用中，主体的决策过程基于模糊逻辑规则，并且根据中央订单簿的基本匹配规则，价格动态是完全确定性的。最后，在将大多数参数置于进化控制之下时，基于计算主体的系统能够复制金融时间序列的几个典型事实（股票收益分布呈现厚尾且具有正超额峰度、股票收益不存在自相关性以及波动聚集现象）。
