资产价格联动的时空演化研究

发布时间：2018-03-17 12:39

本文选题：复杂网络　切入点：股票市场　出处：《南京信息工程大学》2013年硕士论文　论文类型：学位论文

【摘要】：近年来,对中国股票市场行为的研究逐渐成为热点。众所周知,中国股票市场波动性大。一般来说,股票市场的波动行为总是与股票间的复杂关系息息相关。传统的方法是利用多元统计模型对股票之间的波动关联性进行研究,然而在利用这些模型进行研究时往往受到“维数灾难”的限制,较难进行有效的实证。复杂网络理论的诞生方便了我们对资产价格联动特性的研究。本文从系统科学的角度出发,基于复杂网络理论对股票市场中资产价格联动与市场稳定问题进行研究,论文的主要工作和创新成果如下： (1)以上海市场2001年1月2日至2011年3月11日期间的501只股票的日收盘价为样本,采用互信息衡量股票价格之间的相互关系,并采用滑动窗口的方法构建出相应的2063个全连通的上海市场股票动态关联网络。从每个股票网络中提取最大生成树,分析了树的平均路径长度,中心节点影响力和p值随时间的变化情况,又进一步对上海股市不同时期最大生成树的变化特性进行了详细分析。实证结果表明：2005年8月8日、2007年10月17日和2008年12月25日左右的这三段时期是上海股票市场的转折时期,在转折时期,股票市场稳定性变差,最大生成树拓扑结构变得松散,股票节点间的分离程度变大,中心节点的影响力减小,树的结构从星型变为链状,对应的度分布也不再是幂律分布。同时,对最大生成树的单步和多步存活率进行分析,实证发现：在短期的演化过程中,股票间的紧密关系不容易被打破；相反,在长时间的演化过程中,往往不存在一直都是紧密联系的股票节点对。 (2)以2008年12月25日这个股票市场转折点为背景,将它作为转折前期和转折时期的分界点,构建转折前期和转折时期两个行业关联网络,并采用两种方法对网络拓扑特性进行分析。一方面,对原始网络进行阈值化处理,实证发现在转折时期,网络的聚类系数明显减小,凝聚度变差,网络变得松散,进一步通过对网络的k核分解,发现转折时期处于最深核的节点与转折前期相比发生了明显变化,纺织服饰行业和电子行业成为最深核节点,这与当时我国的经济现象十分吻合。另一方面,提取原始网络的最大生成树,实证发现转折时期节点间的分离程度变大,中心节点影响力减小,转折前期与中心节点紧密相连的行业在转折时期全部发生变化。比较以上两种方法,虽然都能揭示行业网络的特性,得到相似的实验结果,但是通过最大生成树来观察行业间联动性的变化则更为直观。以上结论不仅能帮助我们进一步了解股票市场资产价格联动与市场稳定的关系,还能对股票投资风险管理提供一定的指导意义。
[Abstract]:In recent years, the research on the behavior of Chinese stock market has gradually become a hot spot. As we all know, the volatility of Chinese stock market is great. Generally speaking, The volatility behavior of stock market is always closely related to the complex relationship between stocks. However, the use of these models is often limited by "dimensionality disaster", so it is difficult to carry out effective demonstration. The birth of complex network theory facilitates us to study the property of asset price linkage. From the point of view of system science, this paper studies the linkage of asset prices and market stability in stock market based on complex network theory. The main work and innovative results of this paper are as follows:. Using the daily closing price of 501 stocks in the Shanghai market from January 2nd 2001 to March 11th 2011 as a sample, using mutual information to measure the interrelationship between stock prices, A sliding window method is used to construct 2063 fully connected dynamic correlation networks in Shanghai stock market. The maximum spanning tree is extracted from each stock network and the average path length of the tree is analyzed. The influence of the center node and the change of p value over time, Furthermore, the variation characteristics of the largest spanning tree in different periods of Shanghai stock market are analyzed in detail. The empirical results show that the three periods about August 8th 2005, October 17th 2007 and December 25th 2008 are the turning points of Shanghai stock market. In the transition period, the stability of the stock market becomes worse, the topological structure of the maximal spanning tree becomes loose, the degree of separation between the stock nodes becomes larger, the influence of the central node decreases, and the structure of the tree changes from star to chain. The corresponding degree distribution is no longer a power-law distribution. At the same time, the one-step and multi-step survival rates of the maximal spanning tree are analyzed. It is found that the tight relationship between stocks is not easily broken in the short-term evolution process; on the contrary, In the long-term evolution process, there is often no stock node pair which has been closely related. Based on the turning point of the stock market in December 25th 2008, this paper regards it as the dividing point between the early turning period and the turning period, and constructs two related networks of industries in the early turning period and the turning period. Two methods are used to analyze the network topology. On the one hand, the threshold value of the original network is analyzed. It is found that in the transition period, the clustering coefficient of the network decreases obviously, the cohesion becomes worse, and the network becomes loose. Further, by decomposing the k-core of the network, it is found that the nodes in the deepest core in the transition period have changed obviously compared with the earlier transition period, and the textile and clothing industry and the electronics industry have become the deepest core nodes. On the other hand, by extracting the largest spanning tree of the original network, it is found that the separation between nodes becomes larger and the influence of the central node decreases during the transition period. The industries that are closely connected to the central nodes in the early transition period have all changed during the transition period. Comparing the above two methods, although the characteristics of the industry network can be revealed, similar experimental results can be obtained. But it is more intuitionistic to observe the change of inter-industry interaction through the maximal spanning tree. These conclusions can not only help us to understand the relationship between the linkage of asset prices and market stability, but also provide some guidance for stock investment risk management.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51

【参考文献】