论文部分内容阅读
金融数据挖掘是信息社会中一个极具挑战性的研究方向 .金融数据的随机特性使得隐藏在数据中的内在规则难以被发现 .指出了经典相关分析的缺陷 ,进一步讨论了高阶相关系数的性质 ,证明了高阶相关不仅能描述隐藏的非线性相关信息 ,而且正好刻画了线性相关与独立之间的空白 .因此 ,完全可以利用高阶相关性的计算简单性对金融数据中的时变非线性相关特性进行实时跟踪 ,克服了 Brock W .等人于 1987年和 1992年提出的Granger- Causality独立性检验方法中需要正态假设和非实时性的缺点 .最后 ,将上述结果应用于股票价格与成交量之间的相关分析 .数值结果显示高阶相关能跟踪隐藏在数据中的时变非线性相关特性
Financial data mining is one of the most challenging research directions in the information society.The stochastic nature of financial data makes it difficult to find the inherent rules hidden in the data.It points out the defects of classical correlation analysis and further discusses the properties of higher-order correlation coefficients , Which proves that higher-order correlation can not only describe the hidden non-linear related information, but also characterize the gap between linear correlation and independence.Therefore, it is entirely possible to make use of the computational simplicity of higher-order correlation to analyze the time-varying non-linearity in financial data Linear correlation of real-time tracking overcoming the shortcomings of normal and non-real-time required by Brock W. et al.’s Granger-Causality independence test method proposed in 1987 and 1992. Finally, the above results are applied to the stock price And volume.Numerical results show that the high-order correlation can track the time-dependent non-linear related characteristics hidden in the data