A Study Of Reusability-oriented Distributed Multi-agent Financial Market Simulation System

With the development of computing technology,using a computer program to simulate financial markets has become a new way to study the phenomenon showed in financial markets.The combination of complex adaptive system theory and financial field forms the agent-based financial market simulation method.Distributed computing technology enhances the performance of computer simulation.Because of lacking of professional computer knowledge,when researchers in financial field program for a financial market simulation task through general-purpose multi-agent simulation platforms,they may have issues of difficulties in general model modification,parallel program coding,and the issue of high repetition of their developing work.Thus,in this paper,we studied and realized a distributed multi-agent simulation system in the field of financial markets,with regarding the capability of reusability as the most important feature supported by the system.By reusing existing models and codes in this system,it could simplify the progress of constructing a new financial market model,reduce the repetition of system developing work.Four parts of work are included in this thesis.Firstly,from general theories of software reusing,we analyzed the core meaning of realizing the reusability of financial simulation platform.The progress of reusing of financial simulation system,is the progress of the general financial market model deriving to specified financial market models when constructing new financial market simulation programs.Next,we proposed a general financial market model with reusability theoretically in the field of financial market simulation.The variety of market agents,market assets,and communication networks between agents in financial market models is the reason of causing complexity and variety in financial markets,and also the solution of issue of model reusability.The general financial market model proposed in this thesis modeled from those three aspects mentioned above,making it to derivate to any kind of specific financial market experiment models.This work realized the reusability in model level,reducing the difficulty in modeling.Then,through the development process using the reusable component technology,we realized a distributed multi-agent simulation system based on the general model proposed in this thesis.The system adopted an architecture with one simulation middleware and several external parts.The simulation middleware is a running environment for distributed simulation programs,providing services such as message forwarding and simulation synchronizing.External parts realize specific financial market models,plugging into the simulation middleware and running as plugins.Components composing the simulation middleware and external parts could be recomposed again,generating a new version of a financial market model or the middleware.This work realized the reusability in code level,reducing difficulties of coding and repetition of work in the progress of system developing.Finally,we designed experiments to examine the effectiveness and reusability of out simulation system,and it is proved that our system can simulate financial markets in the real world effectively,and has capability of reusing into new financial market models.