金融时间序列的多重分形分类

Multifractal Classification of Financial Time Series

金融时间序列中的多重分形性质经常与标度不规则性和自相似性相联系,经常用多重分形谱方法对金融时间序列进行分类.提出了谱的宽度和峰值的概念,用来区分不同样本的分布特征,用描述统计的方法来得到内在多重分形性的统计描述特征.研究中所用68个证券市场的日交易数据来自纽约股票交易所20 a的交易数据.结果表明,多重分形特征适用于对金融时间序列进行分类.

Multifraetal behavior in financial time series is usually connected with irregular scaling behavior and self-similarity. A multifractal spectrum is used to classify the financial data. The width and the peak of spectrum are proposed as the distinguishing features among the samples. The data, daily series of 68 different assets, coming from New York Stock Exchange over a period of about twenty years, are used in the study. Analysis based on descriptive statistics is provided to obtain the statistical description of the inherited multifractality. The multifractal process is validated to be appropriate in the classification of the financial data.