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:
- Generate multiple random subsamples of the data
- Cluster each subsample
- Measure agreement between clusterings on shared points
- 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. |