Research On Agent-Based Corporate Account Capital Flow System Simulation In Financial Network

Currently, capital flows, including illegal transfers, are gaining much speed and bigger quantity along with the rapid growth of financial informatization and financial innovation. Consequently, online and realtime capital flow supervision becomes increasingly important, which demands a more in-depth research into it. However, the financial system is so huge and complex that it can hardly be approached by existing mathematical model. Instead, Agent-based simulation technology has been widely used in the research of financial area. It builds agents with self-determination and leaning abilities to represent real units in microeconomy system. With the agents communicating with each others like in the real economy, it can simulate a dynamic and complex financial system, and in turn providing tools and possibilities for further studies.In this paper, multi-agent based simulation technology is adopted to study the capital flow in financial system and the capital flow of typical automobile manufacturer in China are choosed as main target research object.An improved BDI model is introduced to facilitate the event-based simulation method. In the next stage, according to the accounting theory, the finance and capital flow of an automobile company are broken down and studied through interviews and surveys. After that, a series of hypersesises are introduced to simplify the model and an agent for car makers is built. It contains several common components and three core modules, namely attribute-set, event-set and rules set. By creating agents with various attributes and rules, the system can simulate the automobile companies in real business circle.Afterwards, the event-based simulation system was designed and developed according to the Agent model. In the system, there are four kinds of participants: corporates, individuals, banks and government. In the progress of simulation, the agents make deals and transfer money with them. Then, 3 parallel experiments are conducted and the results are analyzed to prove its validity and usability. Also, some drawbacks are pointed out and improvements are suggested.In conclusion, in this paper an improved Agent model is introduced, built and developed. The research provided both a feasible approach to the study of complexity of financial system and some valid data which can be used in futher studies.