Modeling And Study Of Muti-agent Artificial Stock Market Based On Imitation Behavior

Stock market is a complex adaptive system. It can not meet the demand by onlyusing pure mathematical theory of modeling method to study stock market. WhenTraditional Finance can not give reasonable explanations to some phenomenons whichoccurred in the financial market, Behavioral Finance has emerged. This theory canexplain microscopic reasons of macroscopic phenomenons by studying and researchinvestors’ cognitive psychology and behavior. In the1990s, people try to use computermodeling method to simulate the reality economic system. And during this time, SantaFe Institution has established an artificial stock market, which is the first organizationfounding the Computing Finance. Ten years later, American professor Holland puttedforward a theory which is "a Complex Adaptive System" giving a new idea to helppeople further understanding the financial market. Beside, this is also a big leap inunderstanding movement and evolution of word. In addition, the theory of "Adaptioncreates complexity" has also become an important basic theory to build artificial financemarket for humanity.Currently, previous studies of modeling artificial stock market have littleconsidered the impact of investors’ psychological factors on investment decisions andthe stock market. In this article, it firstly analyzes the complex adaptive system theory,Behavior Finance, imitation behavior, Muti-Agent modeling method and the Swarmplatform which is a simulation platform. Secondly, it analyzes in details of modeling ofAgent, stock, imitation behavior and environment of stock market. At last, it usedMuti-Agent modeling method established an artificial stock market by imitationbehavior theory on Swarm(Java) platform. The contribution of this article are asfollows:First, this article used the imitation behavior in Behavioral Finance andMuti-Agent modeling method to established an artificial stock market.Second, considering the investor classification problem, this article doesn’t simplydivide them into purely imitational trader or purely rational trader, but uses fuzzymathematics divided investors’ behaviors which based on probability judgment. Suchclassification is more close to the reality of investors’ behavior.Third, the artificial stock market model in this article divides Agent by thoseregions. That is to say, if a agent wants to imitate other agents, it can only imitate it’sneighborhood. This design can avoid imitation aggregation, which is all Agent imitatingonly one or two agents. This is also more close to the reality of stock market.At the end of this article it uses stock market model conducting experiments and byverifying some simulation data such as stock price time series, the distribution ofAgents’ wealth and so on. By designing investors’ imitation probability distribution, the experiment results show that imitation can affect the stability and effectiveness of stockmarket. At the same time, it also conduct experiments about the effects of price limits,no risk assets frequency and genetic algorithm frequency on the stock market. And getthe same results as a reality stock market which can prove some theory of financialmarkets from the view of computer simulation.