A survey on time series anomaly detection
CHEN Furong
XIONG Chen
LI Ting
ZHONG Chao
MA Zhaoyang
LI Da
WANG Jing
Abstract：Time series anomaly detection,however,faces numerous challenges due to the complexity of data characteristics,algorithmic requirements,and diverse application scenarios.To address this,this paper presents a comprehensive survey of time series anomaly detection.First,the paper system-atically analyzes the complexity and challenges of time series anomaly detection tasks from three di-mensions:data characteristics,algorithm requirements,and application scenarios.Second,it catego-rizes anomalies in time series into point anomalies,subsequence anomalies,and inter-variable correla-tion anomalies,providing a detailed exposition of the definitions and detection methods for each type.Third,the paper reviews and analyzes the use of traditional statistical methods,machine learning tech-niques,and deep learning approaches in time series anomaly detection,evaluating their applicability and limitations.Subsequently,it compiles widely used time series anomaly detection datasets,analyz-ing the application scenarios and unique features of each dataset.Finally,it discusses future research directions in time series anomaly detection from five perspectives:anomaly localization,anomaly clas-sification,precursor forecasting,interpretability,and integration with large-scale models.The review highlights that current challenges,including data scarcity,anomaly diversity,and concept drift,re-main unresolved.Future anomaly detection research is expected to evolve toward more granular tasks such as anomaly localization and prediction.
Keywords：time seriesanomaly detectionanomaly patternsdetection algorithms
Publication Date：2025-06-30
Online Publishing Date：2025-08-21(First online date of this platform, not the publication date of the document)
Pages：13( 1-13 )