# Dive into Time-Series Anomaly Detection: A Decade Review

**Authors:** Paul Boniol, Qinghua Liu, Mingyi Huang, Themis Palpanas, John Paparrizos  
**arXiv:** [2412.20512](https://arxiv.org/abs/2412.20512)  
**Date:** 2024-12-29  
**Summary page:** https://www.summarizepaper.com/en/paper/2412.20512v1/

## Key points

- Authors: Paul Boniol, Qinghua Liu, Mingyi Huang, Themis Palpanas, John Paparrizos
- Critical need due to advances in data collection technology and exponential growth of streaming data
- Crucial activity with diverse applications in cyber security, financial markets, law enforcement, and health care
- Evolution from Statistical Measures to Machine Learning Algorithms:
- Growing demand for structured understanding of research methods for time-series anomaly detection
- Presents a survey categorizing and summarizing existing solutions for anomaly detection within a process-centric taxonomy specific to time series context
- Conducts meta-analysis of existing literature to identify general trends in time-series anomaly detection research
- Offers insights into evolving landscape of anomaly detection techniques and methodologies over the past decade

## Summary

In their paper titled "Dive into Time-Series Anomaly Detection: A Decade Review," authors Paul Boniol, Qinghua Liu, Mingyi Huang, Themis Palpanas, and John Paparrizos delve into the critical need for time series analytics in light of recent advances in data collection technology and the exponential growth of streaming data. The focus of their research lies in time-series anomaly detection, a crucial activity with diverse applications across various fields such as cyber security, financial markets, law enforcement, and health care. Traditionally, anomaly detection has been rooted in statistical measures. However, with the proliferation of machine learning algorithms in recent years, there is a growing demand for a structured and comprehensive understanding of research methods for time-series anomaly detection. To address this need, the authors present a survey that categorizes and summarizes existing solutions for anomaly detection within a process-centric taxonomy specific to the time series context. In addition to providing an original classification of anomaly detection methods,the authors conduct a meta-analysis of existing literature to identify general trends in time-series anomaly detection research.By offering insights into the evolving landscape of anomaly detection techniques and methodologies over the past decade,this review contributes to advancing knowledge and understanding in this critical area of data analysis.