Subsequence Anomaly Detection

Subsequence anomaly detection focuses on identifying unusual patterns within longer time series data, aiming to pinpoint both single and recurring anomalies without requiring prior knowledge of their characteristics. Current research emphasizes unsupervised methods, employing graph-based representations and novel algorithms like those based on shapelets, ego-networks, and constraint-based graph clustering to improve accuracy and efficiency. These advancements are significant for diverse applications, enabling more robust analysis of complex time series data across fields like environmental monitoring, healthcare, and genomics, where identifying subtle deviations from normal behavior is crucial.

Papers