HDBSCANOptions<T>

Options & Configuration · AiDotNet.Clustering.Options

Configuration options for HDBSCAN (Hierarchical DBSCAN).

Inheritance: ModelOptions → ClusteringOptions<T> → HDBSCANOptions<T>

For Beginners

HDBSCAN finds clusters at all density levels.

Problems with DBSCAN:

  • Need to pick epsilon (cluster radius) - hard to choose!
  • One epsilon can’t handle both dense and sparse areas

HDBSCAN solution:

  • Try ALL possible epsilon values automatically
  • Build a tree of how clusters merge as density changes
  • Pick the most “stable” clusters from this tree

Benefits:

  • No epsilon parameter needed
  • Finds clusters of varying densities
  • Only need minClusterSize (intuitive parameter)
  • Robust noise detection

How It Works

HDBSCAN extends DBSCAN by constructing a hierarchy of clusters at different density levels, then extracting flat clusters using a stability-based method. It handles clusters of varying densities better than DBSCAN.

Constructors

Constructor Summary
HDBSCANOptions()

Properties

Property Summary
HDBSCANClusterSelection ClusterSelection { get; set; } Gets or sets the cluster selection method.
bool AllowSingleCluster { get; set; } Gets or sets whether to allow single-cluster result.
double Alpha { get; set; } Gets or sets the alpha value for metric.
double ClusterSelectionEpsilon { get; set; } Gets or sets the cluster selection epsilon.
int MinClusterSize { get; set; } Gets or sets the minimum cluster size.
int? MinSamples { get; set; } Gets or sets the minimum samples for core points.