Nowadays, there exist systems in which human and interactions among them play an important role. Due to rapid changes between people and the rules governing the relation ships among them, structure of such systems is becomming move complicated and their analysis is getting harder. In order to examine and evaluate of complex systems, models based on agents can be used and complex systems can be considered as multi-agent systems which their corresponding agents are competing or cooperating. One of these systems in which human and relation among them has a fundamental role is economy and its stronghold i.e. market. Economy is a complex system resulting from many transactions among plenty of agents. Agents of this system are nonhomogeneous and during time, behaviour of them and in large scale behavior of economy system are changing. Introduction of agent-based simulation in financial markets, in which Santa Fe market simulation is one of the leading models, has eased the analysis and evaluation of these markets and their results are approaching to reality more and more. By creating such a simulation, one can analyze different effective factors on these markets using simulation models and then, based on this analysis, implement the necessary decisions in real condition. In this disseration, Santa Fe model is used for modeling stock market and then by applying following changes, stock market is analyzed: changes in learning rules, application of three kinds of independent agents, using different combinations in the model and also applying Genetic algorithm(GA) as the process of learning of agents. Obtained results show that application of learning process causes the market to be move stable than the market without learning. Also the results of other experiments demonstrate that with increase in the kinds of agents in stock market, forcasting of agents will be influenced by other agents, which leads to fluctuations in their forcasting such that the markets which are simulated by the same agent will have the best performance from price forcasting and profit of the next period points of view. Keywords Agent-based systems, stock market, forcasting, Genetic algorithm, learning
