Anomalies in Time Series (ATS)

This repository provides clean, ready to use interfaces for common anomaly detection techniques in time series data, as well as novel implementations.

It also offers a synthetic benchmark to be used more as testbed for understanding each method pros and cons (since benchmarking the unknown is mostly useless [1])

Example notebooks are as well provides, and can be found int he root of the repository.

Supported methods

This is a list of the various methodologies available, together with their capabilities and class/module names. Each also states if it can work on single series (univariate or multivariate), lists of series (univariate or multivariate), or both.

---

MinMaxAnomalyDetector

• Description: trivial anomaly detector based on the minimum and maximum values of the time series
• Type: unsupervised
• Works on: single series (uv or mv)
• Real time: no
• Module: anomaly_detectors.naive.minmax

---

IFSOMAnomalyDetector

• Description: anomaly detector based on Isolation Forest and Self-organizing maps [3]
• Type: unsupervised
• Works on: lists of series (uv)
• Real time: no
• Module: anomaly_detectors.stats.ifsom

---

COMAnomalyDetector

• Description: Robust anomaly detection technique based on the covariance matrix [2]
• Type: unsupervised
• Works on: single series (mv) or lists of series (uv)
• Real time: no
• Module: anomaly_detectors.stats.robust

---

NHARAnomalyDetector

• Description: Robust anomaly detection technique based on non-linear regression via neural networks and residuals modeling [2]
• Type: unsupervised
• Works on: single series (mv) or lists of series (uv)
• Real time: no
• Module: anomaly_detectors.stats.robust

---

Usage

Setup (with virtualenv):

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Run tests:

python -m unittest discover

Optionally:

pip install jupyterlab==4.4.3
jupyter lab --port 9999

References

[1] Robust anomaly detection for time series data in sensor-based critical systems. Stefano Alberto Russo https://arts.units.it/handle/11368/3107341

[2] Anomaly Detection in High-Dimensional Bank Account Balances via Robust Methods. Federico Maddanu, Tommaso Proietti, Riccardo Crupi https://arxiv.org/abs/2511.11143

[3] Navigating AGN variability with self-organizing maps. Ylenia Maruccia, Demetra De Cicco, Stefano Cavuoti, Giuseppe Riccio, Paula Sánchez-Sáez, Maurizio Paolillo, Noemi Lery Borrelli, Riccardo Crupi, Massimo Brescia https://www.aanda.org/articles/aa/pdf/2025/07/aa53866-25.pdf