集群环境下基于并行多主体的股票市场仿真平台研究

发布时间：2018-02-11 10:30

本文关键词： 股票市场仿真基于主体建模并行主体集群环境　出处：《天津大学》2014年硕士论文　论文类型：学位论文

【摘要】：股票市场是金融市场的重要组成部分,建设股票市场仿真平台将为金融领域的研究者提供更可靠的建模与实验手段,可以为金融市场的创新及其风险管理提供“先验性的”模拟仿真,形成金融市场当中政策创新的实验床,为健康的金融创新及其风险管理提供可靠有效的技术手段。现实中的股票市场的参与者是大量以投资者身份出现的个体,这些个体以异质的自主的方式追求更好的市场绩效。在此基础上股票市场可以看作由大量具有适应性并且相互交互的个体组成的,系统结构具有内生演化性的复杂自适应系统。基于主体的建模方法(,)已经成为研究复杂自适应系统的重要方法,本文将采用方法实现一个高效的股票市场仿真平台。传统的仿真方法中,主体之间串行运行,以时间步的方式仿真现实系统中的时间点进而模拟现实主体的并发性。本文在传统方法的基础上引入了并发主体的概念,利用多线程的并发优势实现主体间的并发运行,利用集群环境的计算资源提高仿真效率;另一方面,本文提出了一种并发调度的方案,使得主体的运行不再受限于调度系统,从而更好的反映现实系统中投资者之间的关系。同时,复杂系统的宏观现象在多主体的相互作用下才能正确的“涌现”出来,因此集群环境下的分布式通信也是仿真平台的重要组成部分。为了保证主体间高效的消息传输,本文采用了集中通信与点对点通信相结合的综合通信结构,借助消息队列中间件的跨平台性,实现了消息传输系统,为上层主体的通信屏蔽了底层的物理环境。为了支持消息系统向上提供接口的一致性以及平台管理的一致性,本文提出将股票交易市场看作是一种复杂的主体,与投资者主体具有相同的消息接口和调度接口。为了支持金融领域的研究者在平台上试验金融方面的算法和策略,平台对外开放了相关自定义接口,包括投资者主体决策算法的定制接口,投资者主体交互网络的定制接口,以及市场主体撮合机制的定制接口。本文最后通过平台所表现的“典型的格式化特征”证明了平台的有效性,并且通过对比试验验证了并行多主体股票仿真平台相对于串行系统的性能提升以及随着仿真规模增大而表现出来的扩展性。
[Abstract]:The stock market is an important part of the financial market. The construction of the stock market simulation platform will provide more reliable modeling and experimental means for the researchers in the financial field. It can provide a "transcendental" simulation for financial market innovation and its risk management, and form an experimental bed for policy innovation in financial markets. Provide reliable and effective technical means for healthy financial innovation and risk management. The participants in the real stock market are a large number of individuals who appear as investors. These individuals pursue better market performance in a heterogeneous and autonomous manner. On this basis, the stock market can be seen as consisting of a large number of adaptive and interactive individuals. Agent-based modeling method has become an important method to study complex adaptive systems. This paper uses the method to realize an efficient stock market simulation platform. In this paper, the concept of concurrent agent is introduced on the basis of traditional method, and the concurrency between agents is realized by using the concurrency advantage of multi-thread. On the other hand, this paper proposes a concurrent scheduling scheme, which makes the operation of the main body no longer limited by the scheduling system. In order to better reflect the relationship between investors in the real system. At the same time, the macro phenomena of complex systems can emerge correctly under the interaction of multiple agents. So the distributed communication in cluster environment is also an important part of the simulation platform. In order to ensure efficient message transmission between agents, this paper adopts a comprehensive communication structure which combines centralized communication with point-to-point communication. With the help of the cross-platform nature of message queue middleware, the message transmission system is implemented, and the underlying physical environment is shielded for the communication of the upper body. In this paper, the stock trading market is regarded as a complex entity, which has the same message interface and scheduling interface as the investor agent. In order to support the researchers in the financial field to test the algorithms and strategies of finance on the platform, The platform has opened up the relevant custom interfaces, including the customized interface of the investor agent decision algorithm, the customized interface of the investor subject interactive network, Finally, this paper proves the effectiveness of the platform through the "typical formatting characteristics" of the platform. The performance improvement of parallel multi-agent stock simulation platform compared with serial system and the expansibility of parallel multi-agent stock simulation platform with the increase of simulation scale are verified by comparison experiments.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.9;F830.91

【相似文献】