Recent advances in data collection technology have made time series analytics more crucial than ever, especially with the rise of streaming data. Time-series anomaly detection plays a key role in various fields like cyber security, finance, law enforcement, and health care. While traditional methods focus on statistical measures, the growing popularity of machine learning algorithms calls for a structured approach to research in this area. This survey provides a comprehensive overview of existing anomaly detection solutions, categorizing them under a process-centric taxonomy specific to time series. The author also offers a unique categorization method and identifies general trends in time-series anomaly detection research.
https://arxiv.org/abs/2412.20512