日内金融高频数据的异常点检测  

Outlier detection in intra-day financial data case

作　　者：张维[1,2] 刘博[1] 熊熊[1] 

机构地区：[1]天津大学管理学院,天津300072 [2]天津财经大学,天津300222

出　　处：《系统工程理论与实践》2009年第5期44-50,共7页Systems Engineering-Theory & Practice

基　　金：国家自然科学基金(70471062);天津市2005年度社科研究规划项目(TJ05-TJ003);教育部新世纪优秀人才支持计划(NCET-07-0605)

摘　　要：金融市场上日内的异常波动可能与重大经济事件相关.因此通过分析高频市场数据的异常与市场内在事件属性及其联系,可以帮助市场参与者更深刻地理解市场动态特性.采用了基于最大模量小波转换(WTMM)的多重分形形式,通过上海证券市场的日内指数价格数据来判断高频时间序列下异常点的存在和定位.结果表明,上述小波方法在检测和定位高频金融时间序列数据中的异常点方面是有效的.Financial market experienced a high degree of fluctuation and volatility. This was supposed to be related to economic events. In this paper we employed the multifractal formalism based on WTMM （wavelet transfer modulus maxima） to test the existence and the location of outlier in high frequency time series. High frequency time series show different distributional characteristics from common low frequency data. On the foundation of empirical analysis of the Shanghai Composite Index data, we drew the conclusion that it is reasonable to incorporate this wavelet arithmetic to analyze the properties of intra-day data. The result suggests applied and useful support for market information research.

关 键 词：金融市场 数据处理 时间序列分析 高频 

分 类 号：F831.5[经济管理—金融学]