基于主体的非均衡人工股市计算模型

对SFI系列基于主体的人工股市计算模型进行了改进，改变了资产定价机制，剔除了均衡定价方法，在学习分类器系统中引入了模糊处理环节，并改变了主体的学习方式，使得人工市场中同时包含具有不同学习速度的主体。基于此构建的新人工股市仿真模型的运行结果具有真实股市数据的形式化特征，通过比较不同学习速度的主体以及零智能主体的表现，发现人工股市计算模型中学习对于主体的财富有决定性作用，学习速度也会影响主体的表现，而且学习存在一个限度问题，过多的累积学习次数反而会削弱主体聚集财富的能力，这在基于主体的人工市场的构建中是一个值得重点关注的方面。

Agent-based artificial stock market with out-of-equilibrium pricing model

This paper describes a modification of the artificial stock market models(ASMs) of the Santa Fe Institute(SFI) involving replacement of the equilibrium pricing model with the Walrasian equation,application of fuzzy theory to the learning classifier system(LCS),and changing the mechanism of the performing genetic algorithm(GA) to allow agents with different learning speeds to coexist in the same market.The modified ASM yields simulation time series with stylized empirical properties,which shows that the new LCS can operate effective decision-making and learning.By comparing the wealth level of different agents(including zero intelligence agents),we find that whether an agent undertakes learning or not,and how fast they perform learning both have a decisive effect on their wealth.Furthermore if there is not a limitation on learning in an artificial stock market,which leads to redundant learning by an agent,this leads to a deterioration in their performance.