一种基于小波的高频数据降噪和跳跃信息准则
发布时间：2018-08-05 20:54
【摘要】：检验高频金融数据跳跃点和研究它的波动性在应用中是必要的,例如衍生品定价和风险管理。虽然近些年学者们提出大量的检验跳跃的方法,但这些方法依赖跳跃点数量已知或多重假设检验。这导致了这些检验方法表现出不稳健性以及在实证研究中检验出虚假的跳跃点。另外降噪算法可以清洗数据并估计系统整体趋势,因此对有跳跃点的高频金融数据做降噪处理也是研究中很重要的一部分。基于局部线性尺度逼近(LLSA)和极大重叠离散小波变换(MODWT),本文提出了基于小波的跳跃信息准则(WJIC),它可以同时对数据降噪和识别跳跃点,并且我们构造得分函数优化选择参数。我们通过模拟实验对比WJIC和其它方法的表现,并且把我们的算法应用到美国全国证券交易商协会自动报价表(NASDAQ)。我们证明了 WJIC得到的估计量有良好的渐近性质,模拟及实证研究表明了 WJIC得到的估计量在数值计算中表现很好。
[Abstract]:It is necessary to examine the jump point of high-frequency financial data and study its volatility in applications such as derivatives pricing and risk management. Although a large number of test methods have been proposed in recent years, these methods depend on the number of hopping points known or multiple hypothesis tests. This leads to the unsoundness of these testing methods and the testing of false jump points in empirical research. In addition, the noise reduction algorithm can clean the data and estimate the overall trend of the system, so it is also an important part of the research to reduce the noise of the high-frequency financial data with jump points. Based on local linear scale approximation (LLSA) and maximal overlapping discrete wavelet transform (MODWT), this paper presents a hopping information criterion (WJIC),) based on wavelet, which can reduce data noise and identify jump points simultaneously, and we construct a score function to optimize the selection of parameters. We compare the performance of WJIC and other methods through simulation experiments, and apply our algorithm to the National Association of Securities Dealers automatic quotation form (NASDAQ). We prove that the estimator obtained by WJIC has good asymptotic property, and the simulation and empirical research show that the estimator obtained by WJIC performs well in numerical calculation.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F831.51
本文编号：2167004
[Abstract]:It is necessary to examine the jump point of high-frequency financial data and study its volatility in applications such as derivatives pricing and risk management. Although a large number of test methods have been proposed in recent years, these methods depend on the number of hopping points known or multiple hypothesis tests. This leads to the unsoundness of these testing methods and the testing of false jump points in empirical research. In addition, the noise reduction algorithm can clean the data and estimate the overall trend of the system, so it is also an important part of the research to reduce the noise of the high-frequency financial data with jump points. Based on local linear scale approximation (LLSA) and maximal overlapping discrete wavelet transform (MODWT), this paper presents a hopping information criterion (WJIC),) based on wavelet, which can reduce data noise and identify jump points simultaneously, and we construct a score function to optimize the selection of parameters. We compare the performance of WJIC and other methods through simulation experiments, and apply our algorithm to the National Association of Securities Dealers automatic quotation form (NASDAQ). We prove that the estimator obtained by WJIC has good asymptotic property, and the simulation and empirical research show that the estimator obtained by WJIC performs well in numerical calculation.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F831.51
【相似文献】
相关期刊论文 前3条
1 戴昌钧;;决策的信息准则[J];管理工程学报;1989年Z2期
2 张明玉，，邬文兵;一种基于改进的Akaike信息准则的相关关系检测的新方法[J];中国管理科学;1998年03期
3 郑士贵;一种基于改进的Akaike信息准则的相关关系检测的新方法[J];管理科学文摘;1999年02期
相关硕士学位论文 前2条
1 徐刚;一种基于小波的高频数据降噪和跳跃信息准则[D];中国科学技术大学;2017年
2 庞博;非正态多元线性模型中的二阶偏差校正的Akaike信息准则[D];中国科学技术大学;2017年
本文编号：2167004
本文链接：https://www.wllwen.com/jingjilunwen/huobiyinxinglunwen/2167004.html
