Clustering

🎯 KMeans

Class: KMeans

Hyperparameter	Values
K	Number of clusters
MaxIterations	Convergence limit
Tolerance	Convergence threshold
Seed	Reproducibility
InitMethod	Random, PlusPlus

Exposes after fit: Centroids, Inertia, Iterations

---

🔍 DBSCAN

Class: DBSCAN

Hyperparameter	Values
Epsilon	Neighbourhood radius
MinPoints	Minimum density

Discovers K automatically. Noise points labeled -1. Exposes NoiseCount.

---

🌳 Agglomerative Clustering

Class: AgglomerativeClustering

Hyperparameter	Values
K	Number of clusters
Linkage	Single, Complete, Average, Ward

Bottom-up hierarchical merging. Exposes Dendrogram.

---

📐 Clustering Evaluators

All evaluators implement IClusteringEvaluator where higher score = better.

Evaluator	Class	Metric	Notes
Silhouette	SilhouetteEvaluator	$s \in [-1, 1]$	Higher = better
Inertia (Elbow)	InertiaEvaluator	$-W$ (negated)	Use RawInertia() for elbow curve
Davies-Bouldin	DaviesBouldinEvaluator	$-DB$ (negated)	Lower DB = better separation
Calinski-Harabasz	CalinskiHarabaszEvaluator	$CH$	Higher = better, fast