基于两类学习模型的多主体人工股票市场研究

Artificial stock market of multi-agents based on two learning models

构造了一个由采取基于惯例的学习模型(Roth-Erev模型)和信念学习模型(分类器系统)的两类主体种群组成的人工股票市场模型,该模型的模拟运行结果表明,学习过程是影响资产市场的价格特征的重要因素,当采取不同学习机制的主体数量比例变化时,价格时间序列会表现出不同的统计特征:随机游走过程或异方差性等"格式化的事实".具有创新性的理性更高的主体比例增加未必导致类似理性预期均衡的出现.

In this paper, a multi-agents pricing model is developed to explain stylized facts of stock market. Two representative adaptive learning models, Roth-Erev model and classifier system are adopted by two pop- ulations respectively in order to optimize their strategies in an artificial stock market. The simulation results suggest that the learning mechanism plays a crucial role in market pricing. The properties of observed prices and returns change from random walk to ＂stylized facts＂, while the proportion of heterogeneous agents which learn following vary models is modified. The more sophisticated and creative agents can not help to reach homogeneous rational expectation equilibrium.