StabilityValidation<T>

Models & Types · AiDotNet.Clustering.Evaluation

Stability-based validation for evaluating clustering quality.

For Beginners

Stability validation asks “Is this clustering reliable?”

If you cluster random samples of your data:

  • High stability: Same clusters appear each time (good!)
  • Low stability: Different clusters each time (concerning!)

A stable clustering is more trustworthy because it’s not sensitive to which exact points are included.

How It Works

Stability validation measures how consistent clustering results are across different random subsamples of the data. A good clustering should be stable

  • similar subsamples should produce similar clusters.

Algorithm:

  1. Generate multiple random subsamples of the data
  2. Cluster each subsample
  3. Measure agreement between clusterings on shared points
  4. High stability indicates a robust clustering

Constructors

Constructor Summary
StabilityValidation(int numSubsamples, double subsampleFraction, int? randomState) Initializes a new StabilityValidation instance.

Methods

Method Summary
StabilityAnalysisResult EvaluateRange(Matrix<T> data, int minClusters, int maxClusters) Evaluates stability across a range of cluster counts to find optimal K.
StabilityResult Evaluate(Matrix<T> data, int numClusters) Evaluates clustering stability for a given number of clusters.