CLARANSOptions<T>
Options & Configuration · AiDotNet.Clustering.Options
Configuration options for CLARANS (Clustering Large Applications based on Randomized Search).
Inheritance: ModelOptions → ClusteringOptions<T> → CLARANSOptions<T>
For Beginners
CLARANS randomly explores different cluster configurations.
Think of it like:
- Start with random cluster representatives (medoids)
- Try swapping representatives with random other points
- Keep changes that improve clustering quality
- Repeat until satisfied
Benefits over K-Means:
- Cluster centers are actual data points (medoids)
- More robust to outliers
- Works with any distance function
Trade-offs:
- Slower than K-Means for large datasets
- Randomized: may find different solutions each run
How It Works
CLARANS is a medoid-based algorithm that uses randomized sampling to efficiently search for good cluster medoids. It’s more scalable than PAM while maintaining the benefit of using actual data points as cluster centers.
Constructors
| Constructor | Summary |
|---|---|
CLARANSOptions() |
Initializes CLARANSOptions with appropriate defaults. |
Properties
| Property | Summary |
|---|---|
int NumClusters { get; set; } |
Gets or sets the number of clusters. |
int NumLocal { get; set; } |
Gets or sets the number of local minima to find. |
int? MaxNeighbor { get; set; } |
Gets or sets the number of local search iterations. |