一种基于模板匹配与隶属度解析的时序异常快速检测方法  

A Fast Time Series Anomaly Detection Method Based on Template Matching and Membership Analysis

作　　者：龚汉鑫 齐金鹏[1] 孔凡书 朱俊俊 曹一彤 GONG Hanxin;QI Jinpeng;KONG Fanshu;ZHU Junjun;CAO Yitong(School of Information Science and Technology,Donghua University,Shanghai 201620,China)

机构地区：[1]东华大学信息科学与技术学院,上海201620

出　　处：《电子科技》2022年第6期1-5,27,共6页Electronic Science and Technology

基　　金：国家自然科学基金(61104154);上海市自然科学基金(16ZR1401300,16ZR1401200)。

摘　　要：传统的数据检测技术在处理大规模医疗数据时,耗时较高且抗干扰能力较弱。针对这些问题,文中应用模板匹配与隶属度解析技术,给出了一种时序数据异常状态的快速检测与分析方法。该方法采用TSTKS算法与滑动窗口理论实现时序数据多突变点快速检测,提取连续多窗口波动特征,构建时序数据的归一化波动向量,对大规模病变信号进行异常状态检测与分析。仿真数据与脑电病变信号分析等实验表明,此方法是一种较为快速、准确的大数据分析与检测方法。Traditional data analysis techniques have some problems in processing large-scale medical data detection,such as high time consumption and weak anti-jamming ability.In order to solve these problems,this study presents a fast detection and analysis method for abnormal state of time series data which uses template matching and membership degree analysis techniques.This method uses TSTKS algorithm and sliding window theory to realize rapid detection of multiple mutation points in time series data,extract continuous multi-window fluctuation characteristics,construct a normalized fluctuation vector of time series data,and perform abnormal state detection and analysis on large-scale disease signals.Simulation data and EEG signal analysis show that this method is a relatively fast and accurate method for big data analysis and detection.

关 键 词：突变点检测 大数据分析 异常检测 滑动窗口 时序数据 波动向量 模板匹配 隶属度分析 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TN99[自动化与计算机技术—计算机科学与技术]