FastABODDetector<T>

Models & Types · AiDotNet.AnomalyDetection.AngleBased

Detects anomalies using Fast Angle-Based Outlier Detection (FastABOD).

Inheritance: ModelBase<T, Matrix<T>, Vector<T>> → AnomalyDetectorBase<T> → FastABODDetector<T>

Implements: IGeneratedParameterRegistrar<T>

For Beginners

FastABOD is an optimized version of ABOD that uses only k-nearest neighbors instead of all points. This makes it much faster while maintaining good detection quality.

How It Works

The algorithm works by:

  1. For each point p, find k nearest neighbors
  2. Compute angles only between neighbor pairs
  3. Calculate angle variance as outlier score

When to use:

  • Large datasets where full ABOD is too slow
  • High-dimensional data
  • When you need faster runtime with slight accuracy trade-off

Industry Standard Defaults:

  • K (neighbors): 10
  • Contamination: 0.1 (10%)

Reference: Kriegel, H., et al. (2008). “Angle-Based Outlier Detection in High-dimensional Data.” KDD.

Constructors

Constructor Summary
FastABODDetector(int k, double contamination, int randomSeed) Creates a new FastABOD anomaly detector.

Properties

Property Summary
int K { get; } Gets the number of neighbors used for detection.

Methods

Method Summary
Vector<T> ScoreAnomalies(Matrix<T> X)
void Fit(Matrix<T> X)