EllipticEnvelopeDetector<T>
Models & Types · AiDotNet.AnomalyDetection.Linear
Detects anomalies using Elliptic Envelope (robust covariance estimation).
Inheritance: ModelBase<T, Matrix<T>, Vector<T>> → AnomalyDetectorBase<T> → EllipticEnvelopeDetector<T>
Implements: IGeneratedParameterRegistrar<T>
For Beginners
Elliptic Envelope fits an ellipse (in 2D) or ellipsoid (in higher dimensions) around the data using robust estimation. Points far from this envelope are anomalies. It’s like drawing the smallest ellipse that contains most of the data.
How It Works
The algorithm works by:
- Estimate robust mean and covariance using Minimum Covariance Determinant (MCD)
- Compute Mahalanobis distance for each point
- Points with large distances are anomalies
When to use:
- Data is approximately Gaussian/elliptical
- You need robustness against outliers in the training data
- Multivariate anomaly detection
Industry Standard Defaults:
- Support fraction: 0.5 (use 50% of data for robust estimation)
- Contamination: 0.1 (10%)
Reference: Rousseeuw, P.J., Van Driessen, K. (1999). “A Fast Algorithm for the Minimum Covariance Determinant Estimator.” Technometrics.
Constructors
| Constructor | Summary |
|---|---|
EllipticEnvelopeDetector(double supportFraction, double contamination, int randomSeed) |
Creates a new Elliptic Envelope anomaly detector. |
Properties
| Property | Summary |
|---|---|
double SupportFraction { get; } |
Gets the support fraction (proportion of data for robust estimation). |
Methods
| Method | Summary |
|---|---|
Vector<T> ScoreAnomalies(Matrix<T> X) |
|
void Fit(Matrix<T> X) |