基于多尺度分析的国际汇率价格预测研究

In contrast to the increasing level of complexity in market structure, the characterization of exchange rate markets have been largely constrained by the traditional approaches focusing on homogeneous time horizons and investment strategies, which failed to outperform the benchmark models in the literature. Given the increasing recognition of heterogeneous market structure in the Heterogeneous Market Hypothesis framework, this project proposes the alternative multiscale based analysis and forecasting methodology that attempts to incorporate the mutliscale data characteristics. This project aims to analyze the exchange rate price behaviors and improve the forecasting accuracy, in the complex international exchange rate markets. In the multiscale analysis framework consistent with the Heterogeneous Market Hypothesis, this projects conducts comprehensive studies on the behaviors and the forecasting of exchange rate prices in both Chinese Yuan and other international exchange rate markets worldwide. The main focuses and contributions of the researches in this project involve the following aspects: 1) This project proposes the multiscale based testing methodoloogy of Effiicent Market Hypothesis, attempting to reveal the market effiicency level at finer scales. 2) This project proposes the integrated statistical testing models under the multiscale analysis frameowrk, to test and analyze the exchange rate prices, correlations and dependency structure. Based on the multiscale analysis framework, this project systematically analyzes the price behaivors, espeically the multiscale characteristics in its price and corrlation movement, in the major exchange rate markets. 3) Based on the Hetergeneous Market Hypothesis and multiscale analysis frameowrk, this project extends the traditional fundamental analysis and technical analysis in the time scale domain. As a result, this porjects proposes the multiscale anlaysis based forecasting models, integrating the wavelet analysis and other signal processing methods, curvelet anlaysis and other multiscale geometric analysis, traditonal econometric models and artificial intelligence technqiues. 4) This project constructs a series business intelligent algorithms and applies them to the analysis and forecasting of major exchange rate markets. Based on the analysis and forecating results, this project provides some policy suggestions. The contributions and implications of the present project is as follows: 1) As for theoretical aspect, this project proposes a series of multiscale analysis based forecasting models, which fills the resesarch gaps in the field of forecasting theories and technqiues. 2) As for practical aspect, models proposed in this project provide the quantitative basis of decision support for the pratical business operations and the government decision making process.

本项目针对在复杂多变的国际汇率市场中如何分析和精确预测汇率价格行为的问题，基于异质市场假说和多尺度分析框架，对人民币及国际汇率价格走势预测进行了系统研究。主要研究内容和创新包括：1)提出在多尺度分析框架下汇率市场的有效性检测；2）提出基于多尺度分析框架的统计测试模型，对汇率价格及相关性变动在时频空间的统计特征进行分析测试；3）对传统基础分析与技术分析在时频空间进行扩展，结合以小波分析为代表的波形分析及以曲波为代表的多尺度几何分析技术、传统计量经济模型和人工智能为基础的集成技术，提出一类基于多尺度分析的汇率预测模型；4) 利用多尺度分析框架及开发的预测技术，对人民币及国际汇率市场价格走势进行分析与预测，为相关部门提供决策支持。本项目的研究意义在于：1）理论上，提出基于多尺度分析预测框架，丰富和发展预测理论与技术；2) 应用上，项目的研究成果为企业生产运作和政府决策提供有益的决策

该项目主要针对以汇率为代表的金融时间序列预测领域的一些难点问题，基于多尺度分析基础理论与方法，在基于小波分析、曲波分析、经验模态分解等一系列多尺度分析模型的多尺度预测理论和模型构建研究中取得了一系列进展与研究成果。主要的研究结果包括：(1) 基于异质市场假说和主流多尺度分析方法，以投资策略和投资周期为区分依据，对汇率市场隐含影响因素的微观结构进行分析；(2) 基于主流多尺度分析方法，结合传统技术分析，开发了多个汇率与能源金融价格预测模型；(3) 基于新兴高维多尺度分析方法，以汇率与能源价格为研究对象，结合多元计量模型，建立了多个基于高维多尺度分析的汇率与金融市场联动关系预测模型，并进行了相应的实证分析。截止目前，该项目已发表12篇期刊论文，其中11篇被SCI/SSCI检索；已发表12篇会议论文，其中9篇被EI Compendia检索，3篇被CPCI检索；获得1项国内学术奖励，项目负责人担任国内外4次学术会议的技术程序委员会委员，参与组织工作，培养硕士生10人。根据研究计划与完成工作，该项目已经完成项目预期目标。