基于并行处理技术的期货市场微观结构研究

发布时间：2018-03-21 17:39

本文选题：并行处理　切入点：高频数据　出处：《复旦大学》2013年硕士论文　论文类型：学位论文

【摘要】：市场微观结构理论是近二十年来发展最快的金融领域之一。本文属于市场微观结构的研究范畴,文章从市场机制、投资者行为和市场质量的角度讨论了中国期货市场微观结构。期货市场机制分析比较了做市商机制和竞价机制的优劣,介绍了中国期货市场风险管理制度；期货投资者行为分析了中国期货市场投资者结构,认为个人投资者比重过高导致了市场的投机性。针对采用高频数据实证研究市场微观结构中的问题,提出采用机群并行处理技术解决计算机处理能力瓶颈,首次详细介绍了并行处理的技术方法。实证研究上,应用计算机群并行处理分析了棉花、沪铜、沪锌和黄金期货的“日历效应”,指出期货价格波动、价差等指标日内呈“L”型波动,并提出“L”型波动源于日间数据量和日内数据量之间的差距；针对分笔数据研究中样本时间间隔不等问题,采用自回归条件久期模型(ACD)分析了棉花期货合约交易集聚性问题,证实期货交易的集聚性,回归结果显示期货交易久期受一阶滞后项影响,比较了EACD和WACD模型的估计结果,认为在本次实证中EACD和WACD模型没有显示出明显的估计效果差异。最后,以微观结构研究和高频数据研究的商业层面应用为出发点,探讨了高频自动交易在中国的发展。
[Abstract]:The theory of market microstructure is one of the fastest growing financial fields in the past two decades. From the angle of investor behavior and market quality, this paper discusses the microstructure of Chinese futures market, analyzes and compares the advantages and disadvantages of market maker mechanism and bidding mechanism, and introduces the risk management system of Chinese futures market. The behavior of futures investors analyzes the structure of investors in China's futures market, and considers that the excessive proportion of individual investors leads to the speculative nature of the market. In this paper, a cluster parallel processing technique is proposed to solve the bottleneck of computer processing capability. The technical method of parallel processing is introduced in detail for the first time. "Calendar effect" of Shanghai zinc and gold futures, pointing out that futures price fluctuation and price difference fluctuate in "L" type within a day, and pointing out that the "L" type fluctuation originates from the gap between the amount of data between days and the amount of data within days. Aiming at the problem of unequal sample time interval in the study of split data, the paper analyzes the agglomeration of cotton futures contracts by using the autoregressive conditional duration model (ACDD), and proves the agglomeration of futures trading. The regression results show that the duration of futures trading is affected by the first order lag term. The results of EACD and WACD models are compared. It is concluded that the EACD and WACD models do not show significant difference in the estimated effects in this empirical study. Based on the commercial application of microstructure research and high frequency data research, this paper discusses the development of high frequency automatic trading in China.
【学位授予单位】：复旦大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F224;F724.5

【参考文献】