RandomSubspaceDetector<T>

Models & Types · AiDotNet.AnomalyDetection.Ensemble

Detects anomalies using Random Subspace ensemble method.

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

Implements: IGeneratedParameterRegistrar<T>

For Beginners

Random Subspace creates multiple detectors, each trained on a randomly selected subset of features. This helps detect anomalies that may only be visible in certain feature subspaces, which is common in high-dimensional data.

How It Works

The algorithm works by:

  1. Generate n_estimators random feature subsets
  2. Train a base detector on each subspace
  3. Combine scores using averaging or voting

When to use:

  • High-dimensional data
  • When anomalies hide in subspaces
  • For robust detection across feature combinations

Industry Standard Defaults:

  • N estimators: 20
  • Max features: sqrt(n_features)
  • Combination: Average
  • Contamination: 0.1 (10%)

Reference: Keller, F., Muller, E., Bohm, K. (2012). “HiCS: High Contrast Subspaces for Density-Based Outlier Ranking.” ICDE.

Constructors

Constructor Summary
RandomSubspaceDetector(int nEstimators, int maxFeatures, double contamination, int randomSeed) Creates a new Random Subspace anomaly detector.

Properties

Property Summary
int MaxFeatures { get; } Gets the maximum features per subspace.
int NEstimators { get; } Gets the number of estimators.

Methods

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