How to install Luminol and handle dependencies?

Install via pip with 'pip install luminol', which requires numpy. The library supports Python 2.7 and 3.6, but ensure your environment matches these versions to avoid conflicts.

Can Luminol detect anomalies in real-time streaming data?

No, Luminol is designed for batch analysis of time series data, as shown in the examples using dictionaries or CSV files. For streaming, you'd need to implement custom batching or use a different library.

Luminol vs Prophet for time series anomaly detection?

Luminol is lightweight and focused on configurable, threshold-free scoring for root cause analysis, while Prophet offers more advanced forecasting and trend detection. Choose Luminol for simple, diagnostic workflows; Prophet for predictive modeling.

How to visualize anomalies detected by Luminol?

Luminol does not include built-in visualization. You need to use external libraries like Matplotlib or Plotly to plot the time series and anomaly scores returned by methods like get_all_scores().

What algorithms does Luminol use for correlation?

It primarily uses the cross_correlator algorithm, which allows time shifts to match offset peaks, with parameters like max_shift_seconds. This is useful for correlating slightly delayed events in system metrics.

Is Luminol suitable for multivariate time series analysis?

Not directly; Luminol focuses on pairwise correlation between two time series. For multivariate analysis, you'd need to run multiple correlations or extend the library, which can be cumbersome for many metrics.

Open-Awesome

luminol

Apache-2.0Python

A lightweight Python library for anomaly detection and correlation in time series data, enabling root cause analysis.

GitHub

1.2k stars218 forks0 contributors

What is luminol?

Luminol is a lightweight Python library for time series data analysis, specializing in anomaly detection and correlation. It detects anomalies in time series data by assigning scores to data points and identifying unusual patterns, and it correlates anomalies with other metrics to help investigate root causes, such as spikes in network latency or system performance issues.

Target Audience

Developers and data analysts working with time series data for monitoring, diagnostics, or root cause analysis in systems like network performance, application metrics, or infrastructure monitoring.

Value Proposition

Developers choose Luminol for its configurable algorithms, threshold-free anomaly scoring, and integrated correlation capabilities, which streamline automated root cause analysis workflows without relying on rigid thresholds or predefined rules.

Overview

Anomaly Detection and Correlation library

Use Cases

Best For

Detecting anomalies in system monitoring metrics like CPU usage or network latency
Correlating time series data to identify relationships between system events
Automating root cause analysis for performance issues in distributed systems
Analyzing time series data without predefined thresholds or value limits
Investigating spikes or drops in application or infrastructure metrics
Building custom monitoring and diagnostic tools for time series data

Not Ideal For

Real-time anomaly detection on streaming data, as it appears batch-oriented with no mentioned streaming capabilities
Projects requiring advanced machine learning or deep learning models for time series, since it offers only basic algorithms like bitmap_detector
Systems needing multivariate anomaly detection across many metrics simultaneously, as it focuses on pairwise correlation

Pros & Cons

Pros

Configurable Algorithm Selection

Supports multiple anomaly detection (e.g., bitmap_detector, derivative_detector) and correlation algorithms (e.g., cross_correlator) with customizable parameters, allowing tailored analysis for specific use cases.

Threshold-Free Anomaly Scoring

Does not rely on predefined value thresholds; instead, it assigns anomaly scores to each data point, enabling flexible identification of unusual patterns without manual cutoff settings.

Root Cause Analysis Workflow

Facilitates automated investigative logic by correlating anomalies with other metrics (e.g., GC, IO, CPU) to rank potential root causes, as highlighted in the library's philosophy for systems monitoring.

Flexible Data Input Formats

Accepts time series data as CSV file paths, dictionaries, or TimeSeries objects, making integration easy with various data sources without extensive preprocessing.

Cons

Limited Algorithm Ecosystem

Only a few basic algorithms are available for anomaly detection and correlation, which may not cover complex scenarios or state-of-the-art methods, potentially requiring custom extensions.

Outdated Python Support

Officially supports Python 2.7, which is deprecated, and while it also supports Python 3.6, this could lead to compatibility issues with modern Python environments and libraries.

Sparse Parameter Documentation

The README admits that some algorithm parameters have vague meanings (e.g., for bitmap_detector), making it harder for users to tune effectively without diving into the source code.

Frequently Asked Questions

Related Projects

Prophet

Tool for producing high quality forecasts for time series data that has multiple seasonality with linear or non-linear growth.

Stars20,365

Forks4,637

Last commit7 days ago

sktime

A unified framework for machine learning with time series

A python library for user-friendly forecasting and anomaly detection on time series.

Stars9,496

Forks1,027