LoOPDetector<T>

Models & Types · AiDotNet.AnomalyDetection.DistanceBased

Detects anomalies using Local Outlier Probability (LoOP).

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

Implements: IGeneratedParameterRegistrar<T>

For Beginners

LoOP improves on LOF by providing a probability score between 0 and 1, making results easier to interpret. A score of 0 means the point is definitely not an outlier, while 1 means it’s definitely an outlier.

How It Works

The algorithm works by:

  1. Compute probabilistic set distance (PLOF) using Gaussian error function
  2. Normalize using local standard deviation of LOF values
  3. Convert to probability using error function

When to use:

  • When you need interpretable probability scores
  • When comparing outlier scores across different datasets
  • Similar use cases as LOF but with better interpretability

Industry Standard Defaults:

  • K (neighbors): 10
  • Lambda: 3 (controls sensitivity)
  • Contamination: 0.1 (10%)

Reference: Kriegel, H., et al. (2009). “LoOP: Local Outlier Probabilities.” CIKM.

Constructors

Constructor Summary
LoOPDetector(int k, double lambda, double contamination, int randomSeed) Creates a new LoOP anomaly detector.

Properties

Property Summary
double Lambda { get; } Gets the lambda parameter (standard deviations for probability).
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)