KMedoidsOptions<T>

Options & Configuration · AiDotNet.Clustering.Options

Configuration options for K-Medoids (PAM) clustering.

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

For Beginners

K-Medoids uses real data points as cluster centers.

K-Means vs K-Medoids:

  • K-Means: Centers are averages (may not be real points)
  • K-Medoids: Centers are actual data points (medoids)

When to use K-Medoids:

  • Data has outliers (medoids are more robust)
  • Need interpretable cluster representatives
  • Distance metric isn’t Euclidean
  • Categorical or mixed data

PAM (Partitioning Around Medoids) is the classic algorithm:

  1. Initialize with random medoids
  2. Assign points to nearest medoid
  3. Try swapping medoids with non-medoids
  4. Keep swaps that reduce total cost
  5. Repeat until no improvement

How It Works

K-Medoids is similar to K-Means but uses actual data points (medoids) as cluster centers instead of means. This makes it more robust to outliers and allows use of any distance metric.

Constructors

Constructor Summary
KMedoidsOptions() Initializes KMedoidsOptions with appropriate defaults.

Properties

Property Summary
KMedoidsAlgorithm Algorithm { get; set; } Gets or sets the algorithm variant.
KMedoidsInit Init { get; set; } Gets or sets the initialization method.
int NumClusters { get; set; } Gets or sets the number of clusters.