基于序列比对方法的金融波动研究

发布时间：2018-01-21 02:13

本文关键词： 序列比对符号时间序列分析 K-近邻法金融市场波动预测　出处：《天津大学》2012年硕士论文　论文类型：学位论文

【摘要】：随着金融市场波动的加剧及其在全球范围内的广泛传播，采用科学的方法对金融市场波动进行分析，对于预测金融市场波动具有重要意义。其中金融计量学方法在金融市场波动方面的研究已取得了丰硕的成果，但对于复杂的非线性经济系统来说，单纯以金融计量学的方法来研究金融波动无法全方位地把握波动的规律，需要利用新的方法从不同的角度来研究波动问题，作为金融计量学研究方法的补充，符号时间序列分析方法及序列比对方法正可以做到这一点。本文引入生物信息学中的序列比对方法及非参数的符号时间序列分析方法，与已有的K-近邻法相结合，提出一种新的金融波动预测方法。利用符号化后的时间序列数据，将比对目标序列与样本序列进行序列比对，通过动态规划算法回溯出高于匹配得分阈值的K条历史子序列，以此作为K-近邻法中的K个最近邻，分别计算各自的权重，从而得到预测结果。以上证综指、深证成指的高频数据为样本，对其价格波动序列进行实证分析；在成交价格波动这个单一变量的基础上，通过合适的符号化方法将两维时间序列转化为一维时间序列，从而扩展到对成交价格波动与交易时间间隔、成交价格波动与成交量等两个变量同时进行预测，以个股万科的超高频数据为样本，进行实证分析，验证了该方法的可行性和有效性。该方法可以捕获时间序列的非线性特征，降低噪声的敏感性，无需确定数据生成过程符合什么模型，，也不用做出数据是否平稳等假设，不仅可以预测具体的波动值，也可预测波动所处的区间，适用范围广泛。本文第一章阐述了对金融市场波动进行研究的背景、意义及国内外研究现状，并提出本文的创新点。第二章概述了两个重要的基础理论，符号时间序列方法及序列比对方法。第三章提出了基于序列比对方法的高频金融波动预测方法及其详细步骤，并以上证综指和深证成指采样间隔为20分钟的高频数据为样本进行了实证分析，对波动时间序列及波动符号序列分别进行了预测。第四章将单变量预测扩展到双变量预测，以个股万科2010年3月份的超高频数据为样本进行了实证分析。第五章对全文的研究工作进行了总结，指出序列比对方法在金融市场的研究中仍有很大的应用、发展空间，并指明了下一步需要进行的研究及改进。
[Abstract]:With the aggravation of the financial market volatility and its wide spread in the global scope, the scientific method is used to analyze the financial market volatility. It is of great significance to predict the volatility of financial market. The research of financial metrology in the aspect of financial market volatility has made a lot of achievements, but for the complex nonlinear economic system. It is necessary to use new methods to study volatility from different angles as a supplement to the research methods of financial metrology, because it is impossible to grasp the law of volatility in all directions by using the method of financial metrology. Symbol time series analysis method and sequence alignment method can do this. In this paper, sequence alignment method and nonparametric symbolic time series analysis method in bioinformatics are introduced, which are combined with the existing K- nearest neighbor method. A new forecasting method of financial volatility is proposed, which compares the target sequence with the sample sequence by using the symbolic time series data. Through the dynamic programming algorithm to trace the K historical sub-sequences above the matching score threshold, as K nearest neighbors in the K-nearest neighbor method, calculate their respective weights, and then get the prediction results. The high frequency data of Shenzhen stock market index is taken as the sample, and the price fluctuation series is analyzed empirically. On the basis of the single variable of transaction price volatility, the two-dimension time series is transformed into one-dimensional time series by proper symbolization method, which extends to the interval between transaction price fluctuation and transaction time. Two variables, such as fluctuation of transaction price and volume of transaction, are predicted at the same time. The UHF data of individual stock Vanke are taken as the sample to carry on the empirical analysis. The method can capture the nonlinear characteristics of time series, reduce the sensitivity of noise, and do not need to determine the model of data generation process. It can not only predict the specific fluctuation value, but also predict the range in which the fluctuation is located, so it can be used in a wide range. The first chapter describes the background, significance and current research situation of financial market volatility, and puts forward the innovation of this paper. Chapter two summarizes two important basic theories. Symbolic time series method and sequence alignment method. In the third chapter, the prediction method of high frequency financial volatility based on sequence alignment method and its detailed steps are proposed. And based on the Shanghai Composite Index and Shenzhen Composite Index sampling interval of 20 minutes of high-frequency data as a sample for empirical analysis. In chapter 4th, univariate prediction is extended to bivariate prediction. Based on the UHF data of Vanke in March 2010, the empirical analysis is carried out. Chapter 5th summarizes the research work of the full text. It is pointed out that the method of sequence alignment still has a great application in the research of financial market, and the space for development is also pointed out, and the further research and improvement are pointed out.
【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;F224

【参考文献】