移动云端多标签数据时间序列异常值检测研究  

Research on outlier detection in time series of mobile cloud multi⁃label data

作　　者：宋弘亮 江涵 张志刚 王凌云 SONG Hongliang;JIANG Han;ZHANG Zhigang;WANG Lingyun(Ningbo Electric Power Design Institute Co.,Ltd.,Ningbo 315000,China)

机构地区：[1]宁波市电力设计院有限公司,浙江宁波315000

出　　处：《电子设计工程》2025年第6期81-84,90,共5页Electronic Design Engineering

基　　金：国网江苏省电力有限公司科技项目(J2021025)。

摘　　要：由于移动云端多标签时间序列数据存在复杂度较高、噪声干扰大、数据不完整等问题,使得异常值检测损失量过多。为此,研究移动云端多标签数据时间序列异常值检测方法。利用时间序列规划移动云端多标签数据,采用机器学习方法,获取数据时间序列集合,分析随机时间点的时序数据与相邻时序数据潜在向量描述的相互关系,根据持续编码损失函数,实现数据编码。计算拟合数据分布调整的最低消耗,输入到自回归模型实现异常值检测。实验结果表明,研究方法的检测结果低于损失函数阈值10%,具有极高的稳定性。Due to the high complexity,high noise interference and incomplete data of multi-label time series in mobile cloud,the loss of outlier detection is excessive.To this end,the mobile cloud multi-label data time series outlier detection method is studied.Using time series planning mobile cloud multi-label data,machine learning method is adopted to obtain data time series set,analyze the relationship between timing data of random time points and the potential vector description of adjacent timing data,and realize data coding according to the continuous coding loss function.The minimum cost of fitting the data distribution adjustment is calculated and input into the autoregressive model for outlier detection.The experimental results show that the detection results of the proposed method are lower than 10%of the loss function threshold,and the detection process has high stability.

关 键 词：移动云端 多标签数据 时间序列 异常值检测 

分 类 号：TN711.5[电子电信—电路与系统]