我国黄金现货市场分形研究

发布时间：2018-05-10 10:43

本文选题：非线性 + 分形　；参考：《中国海洋大学》2013年硕士论文

【摘要】：作为现代金融研究的理论基础，有效市场假说无法很好地解释市场行为，假设条件也与实际经验不符，这时，非线性科学中的新分支——分形理论应运而生。运用分形理论分析非线性系统中不规则或者不光滑个体，能够较好地解释复杂性系统的特征和行为。集货币、金融、商品属性于一身的黄金近几年一直是国内外投资者和避险者纷纷追逐的对象，黄金市场行为和价格波动也逐渐受到广泛关注。考虑到有效市场理论在分析金融市场时的缺陷，本文借助非线性方法和理论认识我国黄金现货市场、研究我国黄金现货市场波动特征和行为、预测黄金市场价格具有重要的意义和价值。本文的研究工作主要集中在以下几个方面：首先，分析了有效市场假说在解释市场现象中存在的局限和不足，即市场上的参与者并不是完全理性的、收益率不遵循随机游走以及市场价格对信息的非线性反应等，介绍了具有更广泛适用性的分形市场假说，系统地梳理了分形理论和相应的分析方法，分析了研究非线性和复杂系统的分形理论在黄金市场上的应用。其次，判断了数据的平稳性并对其进行平稳化处理，对处理后的收益率序列进行正态性检验、线性检验、非线性检验、R/S分析以及盒子覆盖法分析。研究工作主要有：1）通过ADF、PP等方法综合判断了数据的平稳性，对收益率平稳化处理后的数据进行正态性检验、线性相关检验和非线性相关检验。2）运用重标极差分析法计算黄金日收益率和周收益率序列的Hurst指数，实证结果说明黄金市场是分形的，时间序列存在长期记忆性，并且有112天的非周期循环长度。3）用盒子覆盖法计算时间序列的盒维数，得到的盒维非整数，与前文分析一同从整体上说明了我国黄金现货市场的分形特征。再次，运用多重分形方法继续深入研究黄金现货市场的多重分形特征及其影响因素、原因。研究工作主要有：1）运用多重分形消除趋势波动分析法计算黄金时间序列的广义Hurst指数、标度函数、奇异指数、多重分形谱，结果说明时间序列在消除趋势后，局部结构不一致，存在多重分形特征。2）通过相位随机化和打乱处理，对比原序列和新序列的分形特征，分析影响多重分形的因素。3）分析市场内出现分形结构的原因。最后，，在实证检验了我国黄金现货市场的长期记忆性和标度不变性等分形特征之后，构建分形插值模型，拟合分形插值曲线，并对我国黄金现货市场价格进行预测，为市场投资者和管理者认识市场、预测市场提供了参考。
[Abstract]:As the theoretical basis of modern financial research, the efficient market hypothesis can not explain market behavior well, and the hypothetical conditions are not consistent with practical experience. At this time, a new branch of nonlinear science-fractal theory emerges as the times require. Using fractal theory to analyze irregular or non-smooth individuals in nonlinear systems can better explain the characteristics and behaviors of complex systems. In recent years, gold, which has the property of currency, finance and commodity, has been pursued by investors at home and abroad, and the behavior and price fluctuation of gold market have been paid more and more attention. Considering the defects of the efficient market theory in analyzing the financial market, this paper studies the fluctuation characteristics and behavior of China's gold spot market by means of nonlinear methods and theories. Forecasting gold market price has important meaning and value. The research work of this paper mainly focuses on the following aspects: Firstly, the limitations and shortcomings of the efficient market hypothesis in explaining the market phenomenon are analyzed, that is, the participants in the market are not completely rational, the rate of return does not follow the random walk and the nonlinear response of the market price to the information, etc. This paper introduces the fractal market hypothesis which has more extensive applicability, systematically combs the fractal theory and the corresponding analysis methods, and analyzes the application of fractal theory to the gold market for the study of nonlinear and complex systems. Secondly, the smoothness of the data is judged and processed, the normality test, the linear test, the nonlinear test and the box coverage analysis are carried out. The research work mainly includes: (1) judging the stability of the data synthetically through ADFU PP and so on, and testing the normality of the data after the rate of return is stabilized. Linear correlation test and nonlinear correlation test. 2) the Hurst index of gold daily yield and weekly rate of return is calculated by the method of rescaled range. The empirical results show that the gold market is fractal, and the time series have long-term memory. And there are 112 days of aperiodic cycle length. 3) the box dimension of the time series is calculated by the box covering method, and the box dimension is not an integer. Together with the previous analysis, the fractal characteristics of the gold spot market in our country are explained as a whole. Thirdly, multifractal method is used to study the multifractal characteristics of gold spot market and its influencing factors. The main research work is: (1) the generalized Hurst exponent, scale function, singular index, multifractal spectrum of prime time series are calculated by multifractal elimination trend fluctuation analysis method. The results show that the local structure of time series is inconsistent after eliminating the trend. There are multifractal features. 2) by phase randomization and scrambling processing, comparing the fractal characteristics of the original sequence with the new sequence, and analyzing the factors affecting multifractal. 3) analyzing the reason of the fractal structure in the market. Finally, after testing the long-term memory and scale invariance of China's gold spot market, a fractal interpolation model is constructed to fit the fractal interpolation curve, and the gold spot market price in China is forecasted. It provides a reference for market investors and managers to understand the market and forecast the market.
【学位授予单位】：中国海洋大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.54

【参考文献】