Imitation of Real Market Dynamics by Construction of Multi-agent Based Evolutionary Artificial Market

In this paper, an adaptive system is proposed which attempts to imitate a real market dynamics by combining together the approaches of studies of historical data and researches of multi-agent artificial market. The proportion of different agents is evolved by genetic algorithm in an artificial double auction market. The purpose of this research is to construct an artificial market which generates the dynamics of real market as similar as possible. The model with heterogeneous agents and the environment with which agents and market interact is complicated but controllable by data mining the optimal proportion of the different agents at the input to the market that generates an output which can fit historical data curve. The simulation results suggest that the system performance is close to the expecting values in the testing with adequate training in advance.