分布式环境中基于多Agent的人工股票市场研究

发布时间：2018-06-23 03:21

本文选题：人工股票市场 + 分布式并行　；参考：《天津大学》2012年硕士论文

【摘要】：以SFI-ASM（圣塔菲人工股票市场）为代表的，基于多Agent技术对股票市场进行建模仿真是研究股票市场性质和规律的主要方法之一，取得了一系列的重要研究成果。但这些人工股票市场仍存在着许多不足，主要体现在两个方面，首先这些人工股票市场都是单机串行的，市场中多Agent基于队列调度运行，Agent投资行为之间是伪并行的，与现实股票市场投资者的完全并行性区别很大；其次它们没有也无法实现多Agent之间的实时消息交互，这就严重限制和影响了多Agent间交互网络对市场行为影响的研究。本文针对这些问题，通过将分布式并行技术和消息中间件技术应用到基于多Agent的人工股票市场研究中，实现了一个基于多Agent的分布式并行人工股票市场——DPASM，解决了上述问题，其中主要的研究和工作包括三个部分：首先，对分布式环境下的Agent及多Agent运行控制结构进行了研究，设计了Agent决策算法库和消息交互接口，分别实现Agent的智能决策能力和消息交互能力，采用全分布式多Agent运行控制结构，使Agent之间具备故障独立性，增加了系统稳定性，并且多Agent之间是完全并行的，符合现实股票市场中投资者之间的关系。其次，研究了股票交易市场的交易机制和结构，分布式环境下，Agent可能在任意时刻提交订单，任意时刻也可能有多个Agent订单，因此交易市场采用连续竞价交易机制，同时利用动态线程池技术实现高负载和高并发能力。最后，研究了多Agent之间的消息交互结构及方法，针对多Agent之间交互网络在市场仿真期间固定不变的特点，采用集中式通信结构和消息队列技术，实现了一个轻量级的、有效的多Agent消息交互中间件——ASMMQ，它采用固定发布/订阅方法，实现多Agent之间实时的、异步的消息交互，且消息交互过程对所有Agent都是透明的。另外，本文还实现了一个人机交互终端，用来对市场属性进行初始化设置，，直观的显示股票市场仿真过程中股票成交价格和成交量的情况，并且可以控制仿真的运行。
[Abstract]:Taking SFI-ASM (Santa Fe artificial stock market) as an example, modeling and simulation of stock market based on multiple Agent technology is one of the main methods to study the nature and law of stock market, and a series of important research results have been obtained. However, these artificial stock markets still have many shortcomings, which are mainly reflected in two aspects. First, these artificial stock markets are serial on a single machine, and there is a pseudo-parallelism between the investment behavior of many Agent agents based on queue scheduling in the market. It is very different from the real stock market investors' complete parallelism. Secondly, they do not and can not realize the real-time message interaction between multiple Agent, which seriously limits and affects the research of the influence of multi-Agent interactive network on market behavior. Aiming at these problems, this paper applies distributed parallel technology and message middleware technology to the research of artificial stock market based on multiple Agent, and realizes a distributed parallel artificial stock market based on multiple Agent, which solves the above problems. The main research and work include three parts: firstly, the Agent and multi-Agent running control structure in distributed environment are studied, and the Agent decision algorithm library and message interaction interface are designed. The intelligent decision ability and message interaction ability of Agent are realized, and the fully distributed multi-Agent running control structure is adopted, which makes Agent have fault independence, increases system stability, and is completely parallel between multiple Agent. The relationship between investors in a realistic stock market. Secondly, the trading mechanism and structure of stock trading market are studied. In distributed environment, the agent may submit orders at any time, and there may be multiple Agent orders at any time, so the trading market adopts the mechanism of continuous bidding. At the same time, dynamic thread pool technology is used to realize high load and high concurrency ability. Finally, the structure and method of message interaction between multiple Agent are studied. Aiming at the fixed characteristics of multi-Agent interactive network during market simulation, a lightweight communication structure and message queue technology are adopted. ASMMQ, an effective multi-Agent message interaction middleware, uses a fixed publish / subscribe method to realize real-time and asynchronous message interaction between multiple Agent, and the process of message interaction is transparent to all Agent. In addition, this paper also implements a human-computer interaction terminal, which is used to initialize the market properties, display the stock transaction price and volume in the stock market simulation process, and can control the operation of the simulation.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F830.91;TP18

【参考文献】