基于GNP的多代理人工股市模型

Agent-based model of artificial stock market by using GNP

结合奥地利学派的经济思想,介绍了一种基于GNP算法的多代理人工股市模型。该模型采用GNP算法来模拟交易个体的行为模式,进化他们的决策规则;同时在设计上强化Agent的异质性,并利用GA算法来优化模型参数。仿真结果表明,GNPASM模型表现出很好的统计性能,能够体现真实股市的一些基本特征。

Combined with the idea of Austrian school of economics, a new multi-agent model for artificial stock market was proposed based on Genetic Network Programming. It focused on applying the GNP（ Genetic Network Programming） approach to emulate investment behavior of agents and evolve their trading rules, Simultaneously, this model enhanced the heterogeneity of agents, and searched for an optimal combination of parameter values bascd on GA（ Genetic Algorithm）. Simulation results confirm the effectiveness of this GNP-ASM model through comparison with empirical statistics.