近似周期时间序列分析及其在程序化交易中的应用

项目名称： 近似周期时间序列分析及其在程序化交易中的应用

项目编号： No.11471120

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 数理科学和化学

项目作者： 吴述金

作者单位： 华东师范大学

项目金额： 60万元

中文摘要： 近似周期时间序列是申请者在研究程序化交易中发现的一种时间序列，其特征为具有周期性波动现象，但是波动周期长度不一定相等。近似周期时间序列大量存在，比如太阳黑子序列。另外，几乎所有与节假日有关的金融或经济日数据都是近似周期时间序列，因为相邻两年的同一个节假日之间时间长度是变化的。 本项目首先给出近似周期时间序列的科学概念界定；其次，给出时间序列的近似周期的识别、提取、消除和检验的方法；复次，对近似周期时间序列进行小波分析研究，主要包括近似周期时间序列的消噪、滤波和突变点的检测，着重探讨小波函数的选择问题；再次，给出近似周期时间序列数据变点的检测方法；最后，把近似周期时间序列的分析理论和变点检测方法应用于程序化交易的策略设计中。 开展本项目研究有望丰富和完善时间序列、小波分析、变点理论和程序化交易的策略设计的理论和方法。

中文关键词： 时间序列分析；近似周期；变点检测；小波分析；程序化交易

英文摘要： Approximately periodic time series is a kind of time series that the applicant found when he researched on program trading. Its main character is that it has perodic phenomena but the length of its period is not any constant. There are a lot of approximately periodic time series, such as the sunspot data. Additionally, almost all daily data on finance or economy, which are effected by some festival or holiday, are approximately periodic time series because the range between two adjacent the same festival or holiday is not any constant. The project plans to do five kinds of work. First, estiablishes a scientific concept on approximately periodic time series. Second, brings forward some methods for identifying, extracting, eliminating and testing approximate period of time series. Third, researches on approximately periodic time series by wavelet analysis, mainly including wavelet denoise, filter and detection for signal singularity, especially selection of wavelet functions. Again,gives some methods for detecting change points of approximately periodic time series. Finally, uses the obtained theories and methods to design strategies of program trading. Research on the project hopefully enriches and improves theories and methods of time series, wavelet analysis, change points and strategy design of program trading.

英文关键词： Analysis of time series;Approximate period;Change point detection;Wavelet analysis;Program trading