基于序变换的数据多维子序列相似性搜索仿真  

Similarity Search Simulation of Multi-Dimensional Subsequences Based on Order Transformation

作　　者：姚红[1] 刘玉洁 王淼 YAO Hong;LIU Yu-jie;WANG Miao(Chengdu College of University of Electronic Scienceand Technology of China,Chengdu Sichuan 611731,China)

机构地区：[1]电子科技大学成都学院,四川成都611731

出　　处：《计算机仿真》2020年第9期464-468,共5页Computer Simulation

基　　金：中国高校计算机教育MOOC联盟项目(B190205);电子科技大学成都学院院级课程设计项目((2019)11号)。

摘　　要：针对传统数据多维子序列相似性搜索方法存在搜索时间较长,搜索准确率较低等问题,提出一种基于序变换的数据多维子序列相似性搜索方法。根据分析序列点整体数据形状变化情况,提取时间序列的主要特征点,并对提取到的主要特征点相对应的子序列实施标记预处理,将子序列转换成空间内的曲线,计算曲线间的欧几里德距离,分析数据多维子序列的性质,通过曲率的分段和原理两种性能,对序列内任意点比较小的邻域中的趋向变化现象进行降噪处理,利用相似性阈值的合理设定完成对子序列相似性搜索。仿真结果表明:所提方法能够降低运算时间,从而缩短搜索时间,提高搜索准确率,具有高效性和准确性。Due to long search time and low search accuracy in traditional methods,a method of similarity search for multi-dimensional subsequences based on ordinal transformation was proposed.After analyzing the overall change of data shape of sequence points,we extracted the main feature points of time series and marked the corresponding subsequences of main feature points after pretreatment.Meanwhile,we transformed the subsequences into the curves in space.Moreover,we calculated the Euclidean distance between the curves and analyzed the properties of multi-di⁃mensional subsequences.Through the segmentation and principle of curvature,the trends and changes in the neigh⁃borhood with small arbitrary points in the sequence was denoised.Finally,we completed the similarity search by rea⁃sonable setting for similarity threshold.Simulation results prove that the proposed method can reduce the operation time,and thus to shorten the search time and improve the search accuracy,so it has high efficiency and high accura⁃cy.

关 键 词：时间序列 滑动窗口技术 欧几里德距离 相似性搜索 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]