yzhao062/pyod
PyOD is a comprehensive Python library for multi-modal anomaly detection, offering 60+ detectors and an agentic workflow for AI agents to drive investigations across various data types.
Awesome AI for Infra › Anomaly Detection
zillow/luminaire
Luminaire is an open-source Python library developed by Zillow for automated, hands-off anomaly detection and forecasting in time series data. It incorporates advanced machine learning techniques to identify anomalous patterns and predict future values, considering correlational and seasonal variations, as well as uncontrollable data fluctuations. The library's workflow is structured around three core components: data preprocessing and profiling, modeling, and configuration optimization. The data preprocessing component handles tasks like missing data imputation, outlier removal from training sets, and data transformation, while also generating profiling information such as historical change points and trend changes. The modeling component allows for training various structural and filtering-based time series models, with options for user-defined or optimized configurations. The configuration optimization component integrates hyperparameter tuning to enable a highly automated anomaly detection process, minimizing the need for manual configuration. Luminaire is designed to monitor both batch and high-frequency streaming time series data, making it suitable for a wide range of operational monitoring use cases. It supports detecting sustained fluctuations in data windows rather than just individual data points, which is particularly useful for streaming applications. The project includes examples for both batch and streaming monitoring workflows, demonstrating its practical application in IT operations and related domains where robust time series anomaly detection is crucial.
https://github.com/zillow/luminaire
Also in Anomaly Detection
datamllab/tods
TODS is a comprehensive automated machine learning system for multivariate time-series outlier detection, providing modules for preprocessing, feature extraction, and a wide array of detection algo...
sintel-dev/Orion
Orion is an open-source machine learning library from MIT's Data to AI Lab, focused on unsupervised time series anomaly detection using various AI-driven pipelines.
khundman/telemanom
Telemanom is a framework using LSTMs and automatic thresholding for unsupervised anomaly detection in multivariate time series data, originally developed for spacecraft telemetry.
MentatInnovations/datastream.io
datastream.io is an open-source framework for real-time anomaly detection in streaming data using Python, Elasticsearch, and Kibana.
Stream-AD/MIDAS
MIDAS is a C++ implementation for real-time anomaly detection in dynamic, time-evolving graphs, designed to identify intrusions, fraud, and fake rating anomalies with high accuracy and speed.
activecm/rita
RITA (Real Intelligence Threat Analytics) is an open-source framework that detects command and control (C2) communication by analyzing network traffic, identifying beaconing, long connections, DNS ...
earthgecko/skyline
Skyline is a real-time anomaly detection and time series analysis system designed for passive monitoring of numerous high-resolution metrics without pre-configured models or thresholds.