Minimal agent based model for financial markets II
Minimal agent based model for financial markets II
复制标题
DOI：
10.1140/epjb/e2009-00029-3
复制
发表时间：
2009-02-01
影响因子：
1.6
通讯作者：
Zaccaria, A.
中科院分区：
文献类型：
--
作者：
Alfi, V.;Cristelli, M.;Zaccaria, A.
我们提出了一个详细的研究论文中介绍的基于代理的模型的统计特性I [Eur. Phys. J. B，DOI：10.1140/epj B/e2009-00028-4]及其对乘法动力学的推广。该模型的目的是考虑最低限度的元素，为理解的起源程式化的事实和他们的自我组织。关键因素是基本面代理人，图表代理人，羊群效应和价格行为。前两个要素对应于市场中稳定和不稳定趋势之间的竞争。羊群行为控制着代理人改变策略的可能性，它是这类模型的关键要素。我们考虑一个线性近似的价格动态，允许一个简单的解释模型的动态，对于许多属性，它是可能的分析结果。广义非线性动力学的结果是非常敏感的参数空间和更难以分析和控制。然而，在这两种情况下，关于程式化事实的性质和自组织的主要结果是非常相似的。非线性动力学的主要特点是波动的增强和程式化事实的更明显的证据。我们还将讨论模型的一些修改，以引入与真实的市场相关的更现实的元素。
We present a detailed study of the statistical properties of the Agent Based Model introduced in paper I [Eur. Phys. J. B, DOI: 10.1140/epjb/e2009-00028-4] and of its generalization to the multiplicative dynamics. The aim of the model is to consider the minimal elements for the understanding of the origin of the stylized facts and their self-organization. The key elements are fundamentalist agents, chartist agents, herding dynamics and price behavior. The first two elements correspond to the competition between stability and instability tendencies in the market. The herding behavior governs the possibility of the agents to change strategy and it is a crucial element of this class of models. We consider a linear approximation for the price dynamics which permits a simple interpretation of the model dynamics and, for many properties, it is possible to derive analytical results. The generalized non linear dynamics results to be extremely more sensible to the parameter space and much more difficult to analyze and control. The main results for the nature and self-organization of the stylized facts are, however, very similar in the two cases. The main peculiarity of the non linear dynamics is an enhancement of the fluctuations and a more marked evidence of the stylized facts. We will also discuss some modifications of the model to introduce more realistic elements with respect to the real markets.