Determine Optimal Number of Clusters (Hybrid Approach) — .determine_optimal

Uses hybrid approach: Elbow method (primary) + Silhouette validation (micro-sample). This avoids memory overflow while maintaining clustering quality.

Usage

.determine_optimal_clusters(
  features_matrix,
  method = "kmeans",
  min_k = 3,
  max_k = 6,
  min_cluster_size = 10,
  nstart = 25,
  seed = 123,
  verbose = FALSE,
  use_sampling = FALSE,
  sample_size = 50000,
  memory_fraction = 0.33,
  k_selection_method = c("hybrid", "elbow", "silhouette")
)

Arguments

features_matrix: Numeric matrix
method: Clustering method (for compatibility, currently only kmeans supported)
min_k: Minimum number of clusters
max_k: Maximum number of clusters
min_cluster_size: Minimum cluster size
nstart: Number of random starts for k-means
seed: Random seed
verbose: Print progress
use_sampling: Whether to use sampling
sample_size: Sample size for silhouette
memory_fraction: Fraction of RAM for distance calculations
k_selection_method: Selection method: "hybrid", "elbow", or "silhouette"
k_range: Range of k to test

Value

List with optimal_k, wss_values, silhouette_values, method_used