基于Hilbert-Huang Transform的股指期货分析

发布时间：2018-03-23 05:21

本文选题：EMD分解　切入点：HHT变换　出处：《吉林大学》2013年硕士论文　论文类型：学位论文

【摘要】：随着经济全球化的不断加深，，中国在各国际经济动荡时期仍然屹立不倒，中国资本融入世界经济的比重越来越高，已经成为世界经济体不可分割的重要部分，中国金融体系的现状严重影响世界经济格局。中国期货市场就是在这样的背景下产生和发展。虽然目前的期货市场从规模到品种都具有局限性，但是我们从活跃的交易市场中不难看出中国的期货市场的发展前景是不可限量的，而且目前中国期货已经开始发挥其套期保值和价格发现的作用了。为了更全面，更好的发展中国经济，从金融市场的各个方面与世界发达国家接轨，我们有必要对中国金融期货的发展趋势和影响因素进行深入探讨。通过对已有的交易信息进行分析，从中找出可用规律，为未来的金融期货市场交易提供可用信息，使得中国的金融期货交易系统更健全、更完善。由于金融数据大多是高频的、非平稳的、非线性的，能够准确分析处理这类过程中数据的数据分析方法又是极为有限的,现在可以使用的数字处理的方法多是针对线性非平稳过程或非线性平稳过程，诸如小波分析、Wagner-Ville、傅里叶频谱分析、多种相平面表示法和时间延迟嵌入法。为了详细考察，由现实世界中非线性、非平稳随机过程中产生的数据，我们迫切地需要新方法。希尔伯特-黄变换HHT（Hilbert-Huang Transform）是美国工程学院院士黄锷等人于1998年提出的一种新型的针对大多数是非线性、非平稳过程的数字信号处理方法。用希尔伯特-黄变换进行股指期货数据分析大致分为以下几个步骤：首先将待分析数据进行归一化处理并剔除交易中不合理数据，然后进行经验模分解生成多个本征模函数，最后基于本征模函数进行希尔伯特变换生成Hilbert谱和Hilbert边际谱。本文主要研究了EMD对中国高频金融数据的有效性和完备性以及基于HHT对股指期货的日内特征分析，日内特征分析是根据股指期货的绝对收益率和成交量分别进行得研究和分析。论文对股指期货收益率进行EMD分解，由于现有EMD分解的MATLAB程序不完全适用于大量高频金融类数据，论文中进行了多次实验并设计了数据重构程序；首先介绍了实验数据的生成原理和实验数据的合理性；然后分别进行了希尔伯特-黄变换，针对逐分钟的金融数据的特点对Hilbert变换的取样参数FS进行了多次实验；最后采用最贴近实际的参数值进行Hilbert变换，此过程首先对EMD分解出的IMF结果进行分析和总结，然后对生成Hilbert谱和Hilbert边际谱进行了分析，总结了股指期货的日内趋势和主要周期性。
[Abstract]:With the deepening of economic globalization, China is still standing in various periods of international economic turmoil. The proportion of Chinese capital into the world economy is increasing, and it has become an inseparable and important part of the world economy. The current situation of China's financial system has seriously affected the world economic pattern. It is against this background that China's futures market has emerged and developed. Although the current futures market has its limitations from scale to variety, But it is not difficult to see from the active trading market that the prospects for the development of China's futures market are limitless, and at present Chinese futures have begun to play their role of hedging and price discovery. In order to be more comprehensive, To better develop China's economy and connect with the developed countries in all aspects of the financial market, it is necessary for us to deeply discuss the development trend and influencing factors of China's financial futures. Through the analysis of the existing trading information, To find out the available rules, to provide the available information for the future financial futures market, and to make the financial futures trading system in China more sound and perfect. Because financial data are mostly high-frequency, non-stationary and nonlinear, data analysis methods that can accurately analyze and process such data are extremely limited. Most of the digital processing methods that can be used now are linear non-stationary processes or nonlinear stationary processes, such as wavelet analysis Wagner-Ville, Fourier spectrum analysis, multi-phase plane representation and time-delay embedding. Data from nonlinear, nonstationary random processes in the real world, We urgently need new methods. Hilbert-Huang transform (HHT(Hilbert-Huang transform) is a new digital signal processing method proposed by American Academy of Engineering academician Huang E et al in 1998, which is mostly nonlinear and non-stationary. The analysis of stock index futures data by Hilbert-Huang transform is divided into the following steps: firstly, the data to be analyzed is normalized and the unreasonable data in the transaction is eliminated, and then empirical mode decomposition is used to generate multiple intrinsic mode functions. Finally, Hilbert transform based on intrinsic mode function is used to generate Hilbert spectrum and Hilbert marginal spectrum. This paper mainly studies the validity and completeness of EMD to Chinese high-frequency financial data and the intra-day characteristic analysis of index futures based on HHT. Intra- day characteristic analysis is based on the absolute return and turnover of stock index futures. In this paper, stock index futures yield is decomposed by EMD. Because the existing MATLAB program of EMD decomposition is not fully applicable to a large number of high-frequency financial data, the paper has carried out many experiments and designed a data reconstruction program. Firstly, the principle of experimental data generation and the rationality of experimental data are introduced, then Hilbert-Huang transform is carried out, and the sampling parameter FS of Hilbert transform is tested several times according to the characteristics of minute by minute financial data. In the end, the Hilbert transform is carried out by using the closest parameter value, and the results of IMF decomposed by EMD are analyzed and summarized firstly, and then the generated Hilbert spectrum and Hilbert marginal spectrum are analyzed. The paper summarizes the intraday trend and main periodicity of stock index futures.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F224;F830.91

【参考文献】