基于V-系统的时间序列跳跃点检测新算法

摘要

一些时间序列中跳跃点往往包含比较重要的信息，对其进行检测和定位对实证分析有着非常重要的意义。本文基于一种完备正交函数系、多小波-V系统，构造了离散正交V变换，提出检测跳跃点的一种新算法DOVT-JDA (Discrete Orthogonal V Translation-Jump Point Detection Algorithm)，并针对存在市场微观结构噪音和跳跃的时间序列做了数值模拟。模拟结果表明本文提出的检测跳的新算法DOVT-JDA不仅行之有效，而且计算简单。%Frequently, the jump point of time series contains very important messages. So, the detection and location of jump is of great significance for the demonstration analysis. Here, we propose a new jump point detection algorithm based on V-system and its discrete orthogonal translation named DOVT-JDA (Discrete Orthogonal V Translation-Jump Point Detection Algorithm). Then, we carry out numerical simulations for the time series encompasses market microstructure noise and jump. The experiments have shown the viability and effectiveness of the new jump detection algorithm.