基于谱聚类的股市结构研究

发布时间：2018-01-12 17:16

本文关键词：基于谱聚类的股市结构研究　出处：《华中科技大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着经济体制改革的不断深化，我国的股票市场也在逐步完善和发展，股票市场已经成为我国证券业乃至是整个金融业至关重要的组成部分。上市公司急速增加，投资者要想了解每一种股票是极不容易的，也不胜其烦。股票价格指数是一些金融机构为了适应这种情况和需要，利用自己的业务知识和熟悉市场的优势编制出来的。然而股票价格指数并不能完全反映市场，因此，本文通过板块分析，希望从宏观的角度更全面地了解股市。近年来，随着数据挖掘技术的发展，利用数据挖掘技术分析股票成为研究的热点。在板块研究中，大都利用传统的聚类方法，本文以2008年12月—2012年3月33个板块指数周收盘价为研究对象，利用数据挖掘中的谱聚类算法将33个板块进行聚类，计算每类板块间的相关性，比较每一类的收益率及涨幅。进而利用k-近邻选取出每个板块的1-近邻，构造一个1-近邻网络，并计算网络的中间中心性，研究网络的拓扑结构。从联系的角度去考察股票，而非孤立的角度去考察股票之间的关系，揭示出股市的结构，，进而对股市的有一个宏观的认识。本文利用谱聚类方法和网络的中间中心性对板块指数进行了分析。结果表明，可以通过了解几个核心板块进而了解整个股市。在本文所研究的时间段内，上证指数、纺织服装、工程建筑、电子信息、商业连锁、深证成指是每一类的核心板块，在同类中具有代表性，可以通过了解某一类的核心板块了解同类中的板块。这些研究成果对投资者了解整个股市结构起到有益的帮助作用，有利于投资者更好地从宏观上把握整个股市。
[Abstract]:With the deepening of economic system reform, China's stock market is gradually improving and developing. The stock market has become a vital part of the securities industry and even the whole financial industry in China. The rapid growth of listed companies makes it extremely difficult for investors to understand every stock. The stock price index is a number of financial institutions in order to adapt to this situation and needs. Using their own business knowledge and familiar with the advantages of the market out of. However, the stock price index can not fully reflect the market, therefore, this paper through plate analysis. I hope to understand the stock market more comprehensively from the macro point of view. In recent years, with the development of data mining technology, the use of data mining technology to analyze stocks has become a research hotspot. This paper takes the closing price of 33 plate indices from December 2008 to March 2012 as the research object and uses the spectral clustering algorithm in data mining to cluster 33 plates. The correlation of each plate is calculated, and the yield and increase of each class are compared. Then the 1-nearest neighbor of each plate is selected by using k-nearest neighbor, and a 1-nearest neighbor network is constructed, and the middle center of the network is calculated. This paper studies the topological structure of the network, examines the relationship between stocks from the angle of connection, but not from an isolated angle, reveals the structure of the stock market, and then has a macroscopic understanding of the stock market. In this paper, the spectral clustering method and the meso-centrality of the network are used to analyze the plate index. The results show that we can understand the whole stock market by understanding several core plates. Shanghai index, textile and clothing, engineering construction, electronic information, business chain, Shenzhen stock index is the core of each type of plate, representative in the same class. We can learn about the same kind of plates by understanding the core sectors of a certain class. These research results can help investors to understand the whole stock market structure. It is good for investors to better grasp the whole stock market from the macro perspective.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.51;F224

【参考文献】