An Agent-Based Economic Simulation Platform With Applications

Modern economist considered economy as a complex adaptive system that is composed of numbers of parallel local-interactive agents. These local-interactions form the macro-economic rules that feedback the decisions of local-interactions. The complexity of the economic system result in the bounded rationality of the economic agent while the adaptability of the economic agent result in the evolution of the economic system. The bounded rationality of the economic agent and the evolution of the economic system constitute the basis theoretical of modern economics. As the development of computation technology, people try to do their reach work about economy on compute by simulation experiments. To get over the problem, micro-simulation research, which simulates the whole economy by totalizing the actions at microscopic level in the economy and tries to understand the whole economy, has been taken place. These experiments to get over the border between micro and macro economics are now tried in the field of "agent-based economic simulation"and some developments have been made. In general, in order to theory we need to sophisticate and expand the model by repeating and piling up numbers of researches. Especially in the field of artificial economy, it is important to build the model with the multiple view-point from many different area related to this topic to create a realistic model, and also those collaboration should be well-organized to be cooperative one another, because the economic society which is the target is built form numbers of different factors and is quite complex. However, as we go deep into the research, we encounter several serious problems that have been making difficult for social scientists to develop the agent-based research further. Firstly, the agent-based research requires expert skills of programming and knowledge of engineering. It is impossible to make all of the social scientists master them. This is why most of the researchers in this area are computer-related scientists and few social scientists are there. Secondly, in many cases, some part of the model that constructs the social simulation is not proved scientifically or might just be an ad hoc hypothesis. In those situations, the researchers have to experiment with a variety of models for each of the unreliable function in the model and raise the reliability step by step. But conventional simulation model cannot support this "constructive approach". Thirdly, the field of our research is changing constantly. But the conventional model is lacking in the power of controlling the simulation, let alone amending and extending the simulation. We know, it is irrational and inefficiency that one new simulation amount to one new model. Finally, to research the economy that is a complex system requires the cooperation of lots of researchers, but the actual simulation models cannot provide effectively sharing module and parallel development. Thereby, the concept of simulation platform is proposed to try to provide a sharable framework for economic simulation. Some of the platform, for example the Swarm simulation system, the Trade Network Game Laboratory (TNG Lab) and the Boxed economic simulation platform (BESP), have gained the notable achievement in the different fields. As economy is an evolutional complex system, its research tools must be last out changing with it. People never and shouldn't be satisfied with the present achievement. When the research went further, we found the deficiencies of the existing platforms: (1) Some of them cannot describe and define the economic agents and their behavior precisely; (2) Some are too abstract to apply them to the whole field of economic simulation experiment; (3) Their targets are too specific to extend further; (4) Their specialization is too intense to be transplanted efficiently; (5) They cannot break away the selective format on the basis of initiating parameters to provide user-defined system for social scientist. Based on these understandings, by analyzing the agent-based economic simulation model and present economic simulation platform deeply, we propose an agent-based economic simulation platform (AESP) that supports agent-based economy simulation form analysis and design of target to execution and control of simulation: (1)First, on the basis of foundation classes that are designed in the platform, model builders can design and define the modeled agent object; (2) Second, according to realistic society and modeling requirement, model builders can describe the relations among the agents and initiate the essential data; (3) Third, user may arrange and initiate the periodic events of executing simulation independently; (4) Last, user can also carry out a certain kind of dynamic control to the execution of the simulation till we get the result of the simulation and finish the analysis.From the viewpoint of software engineering, AESP is implemented by C++ Builder language: the object-oriented language. Combined with theory of "Program to interface"and Abstracted form real world, it (1) provides several basic classes for economic simulation, and opens their attributes, relations and sequences to users; (2) adopts the mode that modularizes the development of the model as well as other modes that reduce the coupling in the program, such as "Model Component Builder"and "Model Composer"; (3) carries out dynamic observation and control by using Thread in system level. Accordingly, AESP provides model builders with a sharable, parallel, extendable, controllable, efficient, dynamic, low-coupling, non-cohesion and constructive framework of model development. From the viewpoint of social science, (1) AESP provides a frame of reference for recognizing the target world; (2) it defines the primitive terms for describing the model; (3) it extends the range of modeling to the whole field of agent-based economic simulation; (4) it offers convenient modeling tools of mathematics, statistics and economics (for example, calculation tools, distribution function, GALCS); (5) user can act as a manager to run a dynamic and independent control, but the mechanism that is irrelative to the social science is transparent for user; (6) it supports agent-based economy simulation form analysis and design of target to execution and control of simulation; (7) its define-based pattern of design accords with the requirement of model the realistic world furthest. The second model is about the analysis of transition economies, with it Just as other projects, after the development of the AESP, the foremost mission is to validate the platform. To do it, we introduce the famous micro-simulation model of the U.S. economy-ASPEN -as the reference. In 1996, SNL developed the prototype of the ASPEN. The ASPEN enable us to study the effect on the different economic parameters (GNP, loan interest rate, home sales and price index) by changing the monetary policy (expansion and contraction). According to the settings of the initialization and parameterization in the prototype, we rebuild an ASPEN. By contract the result of our model with the result of the prototype, we testify the validity of the AESP, what is more, the course of development proves fully that the AESP satisfies the requirement of developing an economic simulation platform: efficient,...