时间序列异常检测是工业界中一个重要的研究领域.当前的时间序列异常检测方法侧重于面向完整的时间序列数据进行异常检测,而没有考虑到包含工业场景中网络异常、传感器损坏等所导致的缺失值的时间序列异常检测任务.文中针对工业场景中更加常见的含缺失值的时间序列异常检测任务,提出了一种基于注意力重新表征的时间序列异常检测算法MMAD(Missing Multivariate Time Series Anomaly Detection).具体来说,MMAD首先将包含缺失值的时间序列数据通过时间位置编码对时间序列中不同时间戳的空间关联进行建模,然后通过掩码注意力表征模块学习不同时间戳之间数据的关联关系并将其表征为一个高维的嵌入式编码矩阵,从而将包含缺失值的多元时间序列表示为不含缺失值的高维表征,最后引入条件标准化流对该表征进行重建,以重建概率作为异常评分,重建概率越小代表样本越异常.在3个经典时间序列数据集上进行实验,结果表明,相比其他基线方法,MMAD性能平均提升了11％,验证了MMAD在缺失值场景下进行多元时间序列异常检测的有效性.
Multivariate Time Series Anomaly Detection Algorithm in Missing Value Scenario
Time series anomaly detection is an important research field in industry.Current methods of time series anomaly detec-tion focus on anomaly detection for complete time series data,without considering the time series anomaly detection task contai-ning missing values caused by network anomaly and sensor damage in industrial scenarios.In this paper,we propose an attention representation-based time series anomaly detection algorithm MM AD(missing multivariate time series anomaly detection)for the more common time series anomaly detection tasks with missing values in industrial scenarios.Specifically,MM AD first models the spatial correlation of different time stamps in time series by time position coding.Then,we build an attention representation module to learn the relationships between different time stamps and represent them as an embedded high-dimensional coding ma-trix,thereby representing the multivariate time series with missing values as a high-dimensional representation without missing values.Finally,we design the conditional normalized flow to reconstruct the representation and use the reconstruction probability as the anomaly score,the lower the probability of reconstruction,the more abnormal the sample.Experiments on three classical time series datasets show that,the average performance of MM AD is improved by 11％comparing with other baseline methods,which verifies the efficacy of MMAD to achieve multivariate time series anomaly detection with missing values.
Multivariate time seriesAnomaly detectionMissing-value scenarioAttention mechanismNeural network
NETL
NSTL
万方数据
