A Class Of Complex Adaptive Systems Model And Simulation Method Of Research

It is very difficult to predict the trend of stock and real estate market, which is as an important component of socio-economic system: investors are eagerly looking forward to make a prediction of tomorrow's price to some extent; scholars have sought to understand how market prices are formed. A mathematical model with explicit objective function is set up, and by maximizing the objective function to aid to understand and explain a variety of situation by using traditional finance theory through a variety of simplified, such as the way of investors' prediction and decision-making, market mechanisms. However, Mathematical model can only be used to deal with full certainty and a simple linear system. Mathematical model is almost impossible to establish a complete mathematical model of the dynamic complexity system. Even if the model is established, because of too large, which required complex and difficult theory and techniques to solve, time-consuming and laborious, the bad feasibility, and the result would not be able to test. It is unrealistic to simplify complex system into a series of abstract rules and formula, which can not reflect exactly the nature and may have conflicting results. Although some phenomenon is explained by using the mathematical model, many questions unanswered are left, which caused of the complexity of the market. Although the traditional theory has been aware of this question, they put complexity more attributed to the impact of the outside world random information, and with little regard for the possibility of endogenous complexity.Science of complexity puts forward the new challenge to traditional economics theory, it is no longer agree that the economic system is the result from market stability and balance about supply-demand, whereas, it is agree that the economic system is the result of much interaction from the individual which continuously adjust the relationship between each other under instability situation. The complex adaptive system theory proposed by Professor Holland is outstanding among some schools of complexity science. The CAS theory of SFI mainly adopts the computer model simulation to study complex system, whose research work is characterized by a high degree of importance to the application of computer technology to study the problem complexity. Computer model can be directly described by using combination algorithm with conditions and exchange, which can only be described roughly by using partial differential equations. Computer-based model, rather than partial differential, can reflect the combination complexity during the evolution. Through computer simulation approach to study and observe the complex system can duplicate or reproduce the complexity of the realistic system in the computer environment under a certain sense.In accordance with the limitation of traditional research on the stock and real estate market, attempt to substitute new research methods and setup for traditional mathematical modeling framework by applying the latest achievements of complexity science and the interdisciplinary approach in the paper. Trying to apply complex adaptive system theory to the stock and real estate market, the system framework of the stock and real estate market based on multi-agent technology is set up, and the key technologies involved were discussed. In this paper, the major research is as follows:The internal consistency of the stock and real estate system and complex adaptive system is analyzed, which proves that stock and real estate system is a complex adaptive system according to complex adaptive system theory. Therefore, to certain extent, it is scientific and feasible to adopt complex adaptive system theory in stock and real estate system research. Based on complex adaptive system theory, the characteristics of stock and real estate system are analyzed, pointed out that complexity is the essence of its internal characteristics, the complexity implies the nonlinear effects between the heterogeneous investors and environment. The analysis of the system made by using Efficient Market Hypothesis, traditional math tools (linear, fixed points, differential equation systems) and traditional balanced economics theory is mostly based on certain assumptions, obviously subjective. Therefore, those theories essentially can not describe the complex and changeable character of such a system. Consequently, complex system theory and method must be adopted to understand and research such a system, take a look at what can be found. In view of this, the stock and real estate market are considered as a class of complex adaptive system, and the complex adaptive system modeling and simulation is used to research such system in the paper. At the same time, the feasibility of using complexity theory to model and simulate a category complex adaptive system is proved by expounding the ASM model in brief.The core concept of the complex adaptive system theory is agent. The theory and technology of agent and multi-agent provide a new way for complex system modeling and simulation. For the application of MAS to artificial society, which is still belong to the initial stage, therefore, it is significance of theory exploring to apply MAS to the stock and real estate system. The methods and process of modeling is discussed in the paper: first to identify suitable micro-individual, and to build their model; then set up the interaction between the individuals; finally make up model of the entire system at the macro, which focuses on the design method of agent and interaction between agents. At the same time, according to practical operation of the stock and real estate, such as a class of complex adaptive systems, the multi-agent modeling is applied to research the stock market, real estate market in the paper.During identify suitable micro-individual, the diversity of agents is a significant feature in complex adaptive system, therefore, for the different types of traders, four types of agents are introduced in the paper. Modeling of the micro-individual is the most important design issues in the multi-agent modeling, that is, set up the properties and act rules of each category agent. In the process of designing agent, the issue is decision-making, the objective function, heterogeneity, and study. Such as neural network agents, attempt to substitute neural network for agent, simulate the adaptability of agent by neural network; as far as BP-CT neural network agent, through the neural network approach to the self-development of consistency in agent behavior with CT method to design it's model, the changing goals are got by using of CT method, to train neural network from conduct and effect of the conduct in order to produce internal consistency, which is a reflection of agent cognitive ability.The interaction agent is the key issue of modeling, because the interaction between agents related to the emergence results of the overall model, therefore, in the modeling process, it is necessary to consider the independence of agent, but also consider the interaction between agents. To this end, the ERA scheme is introduced, the model of a class of complex adaptive system is built based on the ERA scheme by me, the model keeps both the environment, which models the context by means of rules and general data, and the agents, with their private data, at different conceptual levels. Some rule-master and rule-maker are set up in the model, rule-master controls agent behavior, and the rule-maker modifies or generates rules, a rule-maker can be used by a number of rule-master (such as the applying of a learning result). The schedule of the model design not only ensures the modularity of the program, but also guarantees the program scalability: When required to add new type of agent, simply modify or add the corresponding rule-master and rule-maker. At the same time, in the model, in order to better simulate the real market a special kind of agent, book is introduced, which is responsible for receiving and implementation of buy and sell orders. The book works on the basis of two matrixes containing sell order in increasing order or buy order in decreasing order. All types of agent submit orders to the book agent, which deal with the orders and carry out the interaction between different types of the agents at the same time. The skill overcomes the subjective act of traditional stock market research, which introduces analysts to intervene the market. More importantly, many stocks may be introduced through the book agent, each book on behalf of each stock, thereby overcoming the deficiencies of single stock in the traditional stock market study.Swarm is a simulation toolkit, which integrates CAS theory, the idea of object-oriented and distributed artificial intelligence technology. With the great power of Swarm, one can simulate the model abstracted from the real economy system, to observe the operating conditions and trends, so as to provide the decision-making of the real world. In the paper, simulation techniques and methods of a category complex adaptive system under the Swarm environment is explored, an example of stock market simulation program is developed based on swarm simulation toolkit through the simulation program, and the complexity of a class complex adaptive system is researched. The simulation research of the actual operation of the stock market is carried out through the design of various types of agent, the dynamic evolution of the stock market under uncertain environment is simulated, many situations is reproduced, which indeed exist, but is difficult to obtain through traditional research methods. Therefore, the feasibility and validity of applying complexity theory and method to the stock and real estate market is verified. The simulation research provides investors a reference for decision-making, at the same time, has a certain referential significance for the study of similar complex systems in terms of researchers. Aim to understand the dynamic specialty of a class of complex adaptive system deeply, instead of try to predict.